> ## Documentation Index > Fetch the complete documentation index at: https://platform.minimax.io/docs/llms.txt > Use this file to discover all available pages before exploring further. # Text to Speech (T2A) WebSocket (Bidirectional) > WebSocket text-to-speech with streaming text input: send text character by character and let the server buffer it into sentences. ## Differences from `/ws/v1/t2a_v2` This API targets **streaming text input** — piping an LLM's streaming output straight into speech, token by token. | | [`/ws/v1/t2a_v2`](/docs/api-reference/speech-t2a-websocket) | `/ws/v1/t2a_v2_bidi` (this API) | | -------------------------------- | ------------------------------------------------------ | ---------------------------------------------- | | Sentence buffering | **Client** must detect sentence boundaries | **Server** buffers automatically | | Sending text char by char | One synthesis per character, choppy audio | Buffered into full sentences first | | Interruption | Not supported, must close the connection | `task_cancel`, and you can continue afterwards | | Sentence boundary events | None | `sentence_start` / `sentence_end` | | `task_finish` | Closes the connection immediately | Flushes the buffer first, then closes | | Flush without ending the session | None | `task_flush` | The voice, audio and pronunciation parameters of `task_start` are **identical** to `/ws/v1/t2a_v2`, so an existing integration can reuse them as-is. ## Event flow 1. Establish the connection and receive `connected_success` 2. Send `task_start` and receive `task_started` 3. Send `task_continue` (**at any granularity, including a single character**); the server buffers the text and starts synthesizing * `sentence_start` is returned when a sentence starts synthesizing * audio arrives in chunks via `task_continued` * `sentence_end` is returned when that sentence is done 4. Send `task_cancel` to interrupt; after `task_canceled` you may keep sending `task_continue` 5. When a turn is over and you want its tail audio right away, send `task_flush`; the session continues after `task_flushed` 6. Send `task_finish`; the server synthesizes any leftover buffered text, then returns `task_finished` and closes the connection Do not conflate the three levels of completion: `is_final` marks the end of audio for one request, `sentence_end` marks the end of the current sentence, and `task_finished` marks the end of the whole session. A connection can host only one synthesis session at a time. Sending `task_start` again after the task has started returns `2206` (illegal event order) and closes the connection. ## Sentence buffering and latency The server detects sentence boundaries from punctuation, so **the punctuation in the text you send directly affects how natural the audio sounds**: | Text situation | Server behaviour | | ------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- | | Ends with sentence-final punctuation (`。！？…!?.` or a newline) | Synthesized **immediately**, no added latency | | Contains secondary punctuation (`，、；：,;:`) | Splits only once enough text has accumulated, so it never emits tiny fragments | | Long text with no punctuation at all | Force-split once it hits the length cap; the boundary is then unrelated to meaning and the audio will sound clipped | | Ends without punctuation, but enough text has accumulated | Flushed after a short silence window | | Ends without punctuation and is short | **Keeps waiting** — the server will not synthesize half a word just to lower latency | The last row has a practical consequence for multi-turn conversations. If a turn ends with text that is both short and unpunctuated, it is not flushed by the short silence window and instead waits for a longer backstop window. Two ways to avoid that wait: **end the last piece with sentence-final punctuation**, or **send `task_flush`** to push it out explicitly. Note that `task_finish` is not a substitute — it closes the connection, so it cannot be used when running multiple turns over one connection. If your application sanitizes an LLM's output before forwarding it, **keep the original punctuation**. Without punctuation the server falls back to splitting purely by length, so pauses land in the wrong places and the audio sounds noticeably unnatural. When the upstream source (for example an LLM) stalls, the audio will contain a matching gap — that is unavoidable. The server will not flush a half-finished sentence to fill the gap, because that would only make the listener hear half a word and then wait anyway. ## Keeping the connection alive A connection that stays idle for roughly 120 seconds is closed by the server with `2201`. "Idle" means the server is neither receiving upstream events nor sending audio. The server **does not send** WebSocket ping frames on its own. If your session has long silent periods (for example while waiting for the user to speak), send pings from the client — the server replies with pong and refreshes the activity timer. Keeping the TCP connection open is not enough to avoid `2201`. ## Send rate and retries `2205` means too much text is queued for synthesis on the server, usually because you are sending faster than synthesis can keep up. `2205` is **not a quota rate limit**, and it is a soft failure: neither the connection nor the session is closed. Just resend that `task_continue` a moment later — **no reconnect and no second `task_start` are needed**. A single `task_continue` longer than 10,000 characters returns `2204`; that piece is skipped and the connection and session likewise stay open. ## AsyncAPI ````yaml api-reference/speech/t2a/api/asyncapi-bidi.json t2a_v2_bidi_websocket id: t2a_v2_bidi_websocket title: T2a_v2_bidi_websocket description: '' servers: - id: production protocol: wss host: api.minimax.io bindings: [] variables: [] address: /ws/v1/t2a_v2_bidi parameters: [] bindings: [] operations: - &ref_3 id: sendMessage title: Send message description: '' type: receive messages: - &ref_5 id: send_task_start contentType: application/json payload: - name: Task Start Event description: >- Sending the `"task_start"` event officially begins the speech synthesis task. The task is considered successfully started when the server returns a `"task_started"` event. Only after receiving this event can you send `"task_continue"` or `"task_finish"` events to the server. type: object properties: - name: event type: string description: >- Controls the instruction being sent. For this step, set to `task_start`. enumValues: - task_start required: true - name: model type: string description: >- The model version to request. Options: `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo`, `speech-01-hd`, `speech-01-turbo`. enumValues: - speech-2.8-hd - speech-2.8-turbo - speech-2.6-hd - speech-2.6-turbo - speech-02-hd - speech-02-turbo - speech-01-hd - speech-01-turbo required: true - name: voice_setting type: object required: true properties: - name: voice_id type: string description: "The ID of the target voice. \r\n- To apply mixed voices, configure the `timbre_weights` parameter and leave this value empty. \r\n- Supports system voices, cloned voices, and AI-generated voices. Below is a selection of the latest system voices (IDs). The full list of available voices can be viewed on the [System Voice ID List](/faq/system-voice-id) or retrieved programmatically using the [Get Voice API](/api-reference/voice-management-get). \r\n - Chinese:\r\n - moss_audio_ce44fc67-7ce3-11f0-8de5-96e35d26fb85\r\n - moss_audio_aaa1346a-7ce7-11f0-8e61-2e6e3c7ee85d\r\n - Chinese (Mandarin)_Lyrical_Voice\r\n - Chinese (Mandarin)_HK_Flight_Attendant\r\n\r\n - English:\r\n - English_Graceful_Lady\r\n - English_Insightful_Speaker\r\n - English_radiant_girl\r\n - English_Persuasive_Man\r\n - moss_audio_6dc281eb-713c-11f0-a447-9613c873494c\r\n - moss_audio_570551b1-735c-11f0-b236-0adeeecad052\r\n - moss_audio_ad5baf92-735f-11f0-8263-fe5a2fe98ec8\r\n - English_Lucky_Robot\r\n\r\n - Japanese:\r\n - Japanese_Whisper_Belle\r\n - moss_audio_24875c4a-7be4-11f0-9359-4e72c55db738\r\n - moss_audio_7f4ee608-78ea-11f0-bb73-1e2a4cfcd245\r\n - moss_audio_c1a6a3ac-7be6-11f0-8e8e-36b92fbb4f95" required: true - name: speed type: number description: |- Speech speed. Higher values result in faster speech. Range: `[0.5, 2]` (default: 1.0). required: false - name: vol type: number description: |- Speech volume. Higher values increase loudness. Range: `(0, 10]` (default: `1.0`). required: false - name: pitch type: integer description: |- Speech pitch adjustment. Range: `[-12, 12]` (default: `0`, original pitch). required: false - name: emotion type: string description: "Emotion control for synthesized speech. Supported values: `[\"happy\", \"sad\", \"angry\", \"fearful\", \"disgusted\", \"surprised\", \"calm\", \"fluent\", \"whisper\"]`. \r\n- By default, the model automatically selects the most natural emotion based on text. Manual specification is only recommended when explicitly needed. \r\n- Available for models: `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo`, `speech-01-hd`, `speech-01-turbo`. \r\n- Option `fluent`, `whisper` is only available for models: `speech-2.6-turbo`, `speech-2.6-hd`. `speech-2.8-hd` and `speech-2.8-turbo` do not support `whisper`." enumValues: - happy - sad - angry - fearful - disgusted - surprised - calm - fluent - whisper required: false - name: english_normalization type: boolean description: >- Enable text normalization in English. Improves performance in digit-reading scenarios at the cost of slightly higher latency. Default: `false`. required: false - name: latex_read type: boolean description: >- Enable LaTeX formula reading. Default: `false`. - Only supports Chinese. When this parameter is enabled, `language_boost` will be set to `Chinese`. - Formulas must be wrapped with `$$`. - If the request contains a formula with `"\"`, it must be escaped as `"\\"`. Example: The quadratic formula ![The quadratic formula](https://filecdn.minimax.chat/public/d6f62e9a-cd3f-4f55-a237-257eef531683.png) should be written as `$$x = \\frac{-b \\pm \\sqrt{b^2 - 4ac}}{2a}$$` required: false - name: audio_setting type: object required: false properties: - name: sample_rate type: integer description: >- Specifies the sampling rate of the generated audio. Supported values: `[8000, 16000, 22050, 24000, 32000, 44100]`. Default is `32000`. required: false - name: bitrate type: integer description: >- Specifies the bitrate of the generated audio. Supported values: `[32000, 64000, 128000, 256000]`. Default is `128000`. *Note: This parameter only applies to audio in `mp3` format.* required: false - name: format type: string description: >- Specifies the format of the generated audio. Supported values: `[mp3, pcm, flac, wav, pcmu_raw, pcmu_wav, opus]`. Default is `mp3`. *Note: `pcmu_raw` and `pcmu_wav` are G.711 μ-law encoded (8 kHz sample rate; `pcmu_raw` is headerless raw, `pcmu_wav` is wrapped in a WAV container). `opus` is Ogg/Opus encoded; in streaming mode, audio chunks must be reassembled in arrival order before decoding.* enumValues: - mp3 - wav - flac - pcm - pcmu_raw - pcmu_wav - opus required: false - name: channel type: integer description: >- Specifies the number of audio channels. Supported values: `[1, 2]`. `1` = mono, `2` = stereo. Default is `1`. required: false - name: pronunciation_dict type: object required: false properties: - name: tone type: array description: >- Defines pronunciation rules for specific characters or symbols. For Chinese text, tones are represented numerically: 1 = first tone, 2 = second tone, 3 = third tone, 4 = fourth tone, 5 = neutral tone. Example: `["omg/oh my god"]` required: false properties: - name: item type: string required: false - name: timbre_weights type: object required: false properties: - name: voice_id type: string description: >- The ID of the voice used for synthesis. Must be specified together with `weight`. Supports system voices, cloned voices, and text-to-voice generated voices. The full list of available voices can be viewed on the [System Voice ID List](/faq/system-voice-id) or retrieved programmatically using the [Get Voice API](/api-reference/voice-management-get). required: false - name: weight type: integer description: >- The weight assigned to each voice. Must be specified together with `voice_id`. Supported range: `[1, 100]`. Up to 4 voices can be mixed. A higher weight value increases similarity to the corresponding voice. Parameter Configuration Example: ```json "timbre_weights": [ { "voice_id": "female-chengshu", "weight": 30 }, { "voice_id": "female-tianmei", "weight": 70 } ] ``` required: false - name: language_boost type: string description: >- Controls whether recognition for specific minority languages and dialects is enhanced. Default is `null`. If the language type is unknown, set to `"auto"` and the model will automatically detect it. Supported values: [`Chinese`, `Chinese,Yue`, `English`, `Arabic`, `Russian`, `Spanish`, `French`, `Portuguese`, `German`, `Turkish`, `Dutch`, `Ukrainian`, `Vietnamese`, `Indonesian`, `Japanese`, `Italian`, `Korean`, `Thai`, `Polish`, `Romanian`, `Greek`, `Czech`, `Finnish`, `Hindi`, `Bulgarian`, `Danish`, `Hebrew`, `Malay`, `Persian`, `Slovak`, `Swedish`, `Croatian`, `Filipino`, `Hungarian`, `Norwegian`, `Slovenian`, `Catalan`, `Nynorsk`, `Tamil`, `Afrikaans`, `auto`] Note: The speech-01 and speech-02 series models do not currently support Persian, Filipino, or Tamil. enumValues: - Chinese - Chinese,Yue - English - Arabic - Russian - Spanish - French - Portuguese - German - Turkish - Dutch - Ukrainian - Vietnamese - Indonesian - Japanese - Italian - Korean - Thai - Polish - Romanian - Greek - Czech - Finnish - Hindi - Bulgarian - Danish - Hebrew - Malay - Persian - Slovak - Swedish - Croatian - Filipino - Hungarian - Norwegian - Slovenian - Catalan - Nynorsk - Tamil - Afrikaans - auto required: false - name: voice_modify type: object description: |- Voice effects configuration. Supported audio formats: 1. Non-streaming: mp3, wav, flac 2. Streaming: mp3 required: false properties: - name: pitch type: integer description: >- Corresponds to the “Deepen/Brighten” slider on the official page. Range: [-100, 100]. Values closer to -100 produce a deeper voice, while values closer to 100 result in a brighter tone. ![pitch](https://filecdn.minimax.chat/public/75af719d-e126-4297-b3cb-416f382e04ec.png) required: false - name: intensity type: integer description: >- Corresponds to the “Stronger/Softer” slider on the official page. Range: [-100, 100]. Values closer to -100 create a stronger, more forceful sound, while values closer to 100 yield a softer tone. ![intensity](https://filecdn.minimax.chat/public/14015a81-d9c4-459b-9536-15c511aac6c0.png) required: false - name: timbre type: integer description: >- Corresponds to the “Nasal/Crisp” slider on the official page. Range: [-100, 100]. Values closer to -100 produce a fuller, richer sound, while values closer to 100 generate a crisper tone. ![timbre](https://filecdn.minimax.chat/public/86ab8ff8-896c-4254-b181-017d9d14000e.png) required: false - name: sound_effects type: string description: >- Sound effects. Only one can be applied at a time. Options: `spacious_echo`, `auditorium_echo`, `lofi_telephone`,`robotic` enumValues: - spacious_echo - auditorium_echo - lofi_telephone - robotic required: false - name: subtitle_enable type: boolean description: >- Controls whether subtitles are enabled. Default is `false`. Available for models: `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo`, `speech-01-hd`, `speech-01-turbo`. required: false - name: subtitle_type type: string description: >- Subtitle granularity. Default is `sentence`. Options: - `sentence`: sentence-level timestamps - `word`: word-level timestamps - `word_streaming`: word-level timestamps optimized for streaming enumValues: - sentence - word - word_streaming required: false - name: continuous_sound type: boolean description: >- Controls model-side text segmentation strategy. Only available for `speech-2.8-hd` and `speech-2.8-turbo` models. - `true`: model does not split the text and performs continuous inference to generate audio (better prosody for long text) - `false`: model splits the text and runs concurrent inference for each segment (lower latency) Default is `false`. required: false - name: session_id type: string description: >- Session id generated by the client to identify this synthesis session. Optional; when omitted the server falls back to `connect_id`. It is echoed back unchanged so the client can correlate its own session context required: false headers: [] jsonPayloadSchema: type: object required: - event - model - voice_setting properties: event: type: string enum: - task_start default: - task_start description: >- Controls the instruction being sent. For this step, set to `task_start`. x-parser-schema-id: model: type: string description: >- The model version to request. Options: `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo`, `speech-01-hd`, `speech-01-turbo`. enum: - speech-2.8-hd - speech-2.8-turbo - speech-2.6-hd - speech-2.6-turbo - speech-02-hd - speech-02-turbo - speech-01-hd - speech-01-turbo x-parser-schema-id: voice_setting: type: object required: - voice_id properties: voice_id: type: string description: "The ID of the target voice. \r\n- To apply mixed voices, configure the `timbre_weights` parameter and leave this value empty. \r\n- Supports system voices, cloned voices, and AI-generated voices. Below is a selection of the latest system voices (IDs). The full list of available voices can be viewed on the [System Voice ID List](/faq/system-voice-id) or retrieved programmatically using the [Get Voice API](/api-reference/voice-management-get). \r\n - Chinese:\r\n - moss_audio_ce44fc67-7ce3-11f0-8de5-96e35d26fb85\r\n - moss_audio_aaa1346a-7ce7-11f0-8e61-2e6e3c7ee85d\r\n - Chinese (Mandarin)_Lyrical_Voice\r\n - Chinese (Mandarin)_HK_Flight_Attendant\r\n\r\n - English:\r\n - English_Graceful_Lady\r\n - English_Insightful_Speaker\r\n - English_radiant_girl\r\n - English_Persuasive_Man\r\n - moss_audio_6dc281eb-713c-11f0-a447-9613c873494c\r\n - moss_audio_570551b1-735c-11f0-b236-0adeeecad052\r\n - moss_audio_ad5baf92-735f-11f0-8263-fe5a2fe98ec8\r\n - English_Lucky_Robot\r\n\r\n - Japanese:\r\n - Japanese_Whisper_Belle\r\n - moss_audio_24875c4a-7be4-11f0-9359-4e72c55db738\r\n - moss_audio_7f4ee608-78ea-11f0-bb73-1e2a4cfcd245\r\n - moss_audio_c1a6a3ac-7be6-11f0-8e8e-36b92fbb4f95" x-parser-schema-id: speed: type: number format: float description: |- Speech speed. Higher values result in faster speech. Range: `[0.5, 2]` (default: 1.0). minimum: 0.5 maximum: 2 default: 1 x-parser-schema-id: vol: type: number format: float description: |- Speech volume. Higher values increase loudness. Range: `(0, 10]` (default: `1.0`). exclusiveMinimum: 0 maximum: 10 default: 1 x-parser-schema-id: pitch: type: integer description: |- Speech pitch adjustment. Range: `[-12, 12]` (default: `0`, original pitch). minimum: -12 maximum: 12 default: 0 x-parser-schema-id: emotion: type: string description: "Emotion control for synthesized speech. Supported values: `[\"happy\", \"sad\", \"angry\", \"fearful\", \"disgusted\", \"surprised\", \"calm\", \"fluent\", \"whisper\"]`. \r\n- By default, the model automatically selects the most natural emotion based on text. Manual specification is only recommended when explicitly needed. \r\n- Available for models: `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo`, `speech-01-hd`, `speech-01-turbo`. \r\n- Option `fluent`, `whisper` is only available for models: `speech-2.6-turbo`, `speech-2.6-hd`. `speech-2.8-hd` and `speech-2.8-turbo` do not support `whisper`." enum: - happy - sad - angry - fearful - disgusted - surprised - calm - fluent - whisper x-parser-schema-id: english_normalization: type: boolean description: >- Enable text normalization in English. Improves performance in digit-reading scenarios at the cost of slightly higher latency. Default: `false`. x-parser-schema-id: latex_read: type: boolean description: >- Enable LaTeX formula reading. Default: `false`. - Only supports Chinese. When this parameter is enabled, `language_boost` will be set to `Chinese`. - Formulas must be wrapped with `$$`. - If the request contains a formula with `"\"`, it must be escaped as `"\\"`. Example: The quadratic formula ![The quadratic formula](https://filecdn.minimax.chat/public/d6f62e9a-cd3f-4f55-a237-257eef531683.png) should be written as `$$x = \\frac{-b \\pm \\sqrt{b^2 - 4ac}}{2a}$$` x-parser-schema-id: x-parser-schema-id: VoiceSetting audio_setting: type: object properties: sample_rate: type: integer description: >- Specifies the sampling rate of the generated audio. Supported values: `[8000, 16000, 22050, 24000, 32000, 44100]`. Default is `32000`. x-parser-schema-id: bitrate: type: integer description: >- Specifies the bitrate of the generated audio. Supported values: `[32000, 64000, 128000, 256000]`. Default is `128000`. *Note: This parameter only applies to audio in `mp3` format.* x-parser-schema-id: format: type: string description: >- Specifies the format of the generated audio. Supported values: `[mp3, pcm, flac, wav, pcmu_raw, pcmu_wav, opus]`. Default is `mp3`. *Note: `pcmu_raw` and `pcmu_wav` are G.711 μ-law encoded (8 kHz sample rate; `pcmu_raw` is headerless raw, `pcmu_wav` is wrapped in a WAV container). `opus` is Ogg/Opus encoded; in streaming mode, audio chunks must be reassembled in arrival order before decoding.* enum: - mp3 - wav - flac - pcm - pcmu_raw - pcmu_wav - opus x-parser-schema-id: channel: type: integer description: >- Specifies the number of audio channels. Supported values: `[1, 2]`. `1` = mono, `2` = stereo. Default is `1`. x-parser-schema-id: x-parser-schema-id: AudioSetting pronunciation_dict: type: object properties: tone: type: array items: type: string x-parser-schema-id: description: >- Defines pronunciation rules for specific characters or symbols. For Chinese text, tones are represented numerically: 1 = first tone, 2 = second tone, 3 = third tone, 4 = fourth tone, 5 = neutral tone. Example: `["omg/oh my god"]` x-parser-schema-id: x-parser-schema-id: PronunciationDict timbre_weights: type: object properties: voice_id: type: string description: >- The ID of the voice used for synthesis. Must be specified together with `weight`. Supports system voices, cloned voices, and text-to-voice generated voices. The full list of available voices can be viewed on the [System Voice ID List](/faq/system-voice-id) or retrieved programmatically using the [Get Voice API](/api-reference/voice-management-get). x-parser-schema-id: weight: type: integer description: >- The weight assigned to each voice. Must be specified together with `voice_id`. Supported range: `[1, 100]`. Up to 4 voices can be mixed. A higher weight value increases similarity to the corresponding voice. Parameter Configuration Example: ```json "timbre_weights": [ { "voice_id": "female-chengshu", "weight": 30 }, { "voice_id": "female-tianmei", "weight": 70 } ] ``` minimum: 1 maximum: 100 x-parser-schema-id: x-parser-schema-id: TimbreWeights language_boost: type: string description: >- Controls whether recognition for specific minority languages and dialects is enhanced. Default is `null`. If the language type is unknown, set to `"auto"` and the model will automatically detect it. Supported values: [`Chinese`, `Chinese,Yue`, `English`, `Arabic`, `Russian`, `Spanish`, `French`, `Portuguese`, `German`, `Turkish`, `Dutch`, `Ukrainian`, `Vietnamese`, `Indonesian`, `Japanese`, `Italian`, `Korean`, `Thai`, `Polish`, `Romanian`, `Greek`, `Czech`, `Finnish`, `Hindi`, `Bulgarian`, `Danish`, `Hebrew`, `Malay`, `Persian`, `Slovak`, `Swedish`, `Croatian`, `Filipino`, `Hungarian`, `Norwegian`, `Slovenian`, `Catalan`, `Nynorsk`, `Tamil`, `Afrikaans`, `auto`] Note: The speech-01 and speech-02 series models do not currently support Persian, Filipino, or Tamil. enum: - Chinese - Chinese,Yue - English - Arabic - Russian - Spanish - French - Portuguese - German - Turkish - Dutch - Ukrainian - Vietnamese - Indonesian - Japanese - Italian - Korean - Thai - Polish - Romanian - Greek - Czech - Finnish - Hindi - Bulgarian - Danish - Hebrew - Malay - Persian - Slovak - Swedish - Croatian - Filipino - Hungarian - Norwegian - Slovenian - Catalan - Nynorsk - Tamil - Afrikaans - auto x-parser-schema-id: voice_modify: type: object description: |- Voice effects configuration. Supported audio formats: 1. Non-streaming: mp3, wav, flac 2. Streaming: mp3 properties: pitch: type: integer description: >- Corresponds to the “Deepen/Brighten” slider on the official page. Range: [-100, 100]. Values closer to -100 produce a deeper voice, while values closer to 100 result in a brighter tone. ![pitch](https://filecdn.minimax.chat/public/75af719d-e126-4297-b3cb-416f382e04ec.png) minimum: -100 maximum: 100 x-parser-schema-id: intensity: type: integer description: >- Corresponds to the “Stronger/Softer” slider on the official page. Range: [-100, 100]. Values closer to -100 create a stronger, more forceful sound, while values closer to 100 yield a softer tone. ![intensity](https://filecdn.minimax.chat/public/14015a81-d9c4-459b-9536-15c511aac6c0.png) minimum: -100 maximum: 100 x-parser-schema-id: timbre: type: integer description: >- Corresponds to the “Nasal/Crisp” slider on the official page. Range: [-100, 100]. Values closer to -100 produce a fuller, richer sound, while values closer to 100 generate a crisper tone. ![timbre](https://filecdn.minimax.chat/public/86ab8ff8-896c-4254-b181-017d9d14000e.png) minimum: -100 maximum: 100 x-parser-schema-id: sound_effects: type: string description: >- Sound effects. Only one can be applied at a time. Options: `spacious_echo`, `auditorium_echo`, `lofi_telephone`,`robotic` enum: - spacious_echo - auditorium_echo - lofi_telephone - robotic x-parser-schema-id: x-parser-schema-id: VoiceModify subtitle_enable: type: boolean description: >- Controls whether subtitles are enabled. Default is `false`. Available for models: `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo`, `speech-01-hd`, `speech-01-turbo`. default: false x-parser-schema-id: subtitle_type: type: string description: >- Subtitle granularity. Default is `sentence`. Options: - `sentence`: sentence-level timestamps - `word`: word-level timestamps - `word_streaming`: word-level timestamps optimized for streaming enum: - sentence - word - word_streaming default: sentence x-parser-schema-id: continuous_sound: type: boolean description: >- Controls model-side text segmentation strategy. Only available for `speech-2.8-hd` and `speech-2.8-turbo` models. - `true`: model does not split the text and performs continuous inference to generate audio (better prosody for long text) - `false`: model splits the text and runs concurrent inference for each segment (lower latency) Default is `false`. default: false x-parser-schema-id: session_id: type: string description: >- Session id generated by the client to identify this synthesis session. Optional; when omitted the server falls back to `connect_id`. It is echoed back unchanged so the client can correlate its own session context x-parser-schema-id: x-parser-schema-id: SendTaskStartEvent title: Task Start Event description: >- Sending the `"task_start"` event officially begins the speech synthesis task. The task is considered successfully started when the server returns a `"task_started"` event. Only after receiving this event can you send `"task_continue"` or `"task_finish"` events to the server. example: |- { "event": "task_start", "model": "speech-2.8-turbo", "language_boost": "Chinese", "voice_setting": { "voice_id": "English_expressive_narrator", "speed": 1, "vol": 1, "pitch": 0 }, "pronunciation_dict": { "tone": [ "Omg/Oh my god" ] }, "audio_setting": { "sample_rate": 32000, "bitrate": 128000, "format": "mp3", "channel": 1 }, "session_id": "my-session-001" } bindings: [] extensions: - id: x-parser-unique-object-id value: send_task_start - &ref_6 id: send_task_continue contentType: application/json payload: - name: Task Continue Event description: >- After receiving the `task_started` event from the server, the task officially begins and you can send `task_continue` events carrying the text to synthesize. Key difference from `/ws/v1/t2a_v2`: **text is not synthesized one message at a time**. It enters a server-side sentence buffer and is only sent for synthesis once it forms a sentence, so it is safe to send text character by character (or token by token): - flushed immediately on sentence-ending punctuation (`。！？…；!?.` and newline) - flushed on secondary punctuation (`，、：,;:`) once the buffered segment is long enough - force-split when the buffer reaches its length limit - after you stop sending for a short while, the leftover buffer is flushed automatically Whitespace-only text is discarded silently and returns no error. If no new event is sent within 120s after the last server response, the WebSocket connection closes automatically. type: object properties: - name: event type: string description: >- The type of session event. For this step, set to `task_continue`. enumValues: - task_continue required: true - name: text type: string description: >- Text to synthesize, limited to fewer than 10,000 characters. **May be sent at any granularity** (including a single character) — the server handles sentence buffering, so the client does not need to detect sentence boundaries itself. - Use newline characters to mark paragraph breaks - Pause control: insert `<#x#>` where `x` is the pause length in seconds, range **[0.01, 99.99]** with at most two decimals. A pause marker must sit between two pronounceable pieces of text; consecutive markers are not allowed - Paralinguistic tags: only supported when the model is `speech-2.8-hd` or `speech-2.8-turbo`. Supported tags: `(laughs)`, `(chuckle)`, `(coughs)`, `(clear-throat)`, `(groans)`, `(breath)`, `(pant)`, `(inhale)`, `(exhale)`, `(gasps)`, `(sniffs)`, `(sighs)`, `(snorts)`, `(burps)`, `(lip-smacking)`, `(humming)`, `(hissing)`, `(emm)`, `(sneezes)` required: true headers: [] jsonPayloadSchema: type: object required: - event - text properties: event: type: string description: >- The type of session event. For this step, set to `task_continue`. enum: - task_continue default: - task_continue x-parser-schema-id: text: type: string description: >- Text to synthesize, limited to fewer than 10,000 characters. **May be sent at any granularity** (including a single character) — the server handles sentence buffering, so the client does not need to detect sentence boundaries itself. - Use newline characters to mark paragraph breaks - Pause control: insert `<#x#>` where `x` is the pause length in seconds, range **[0.01, 99.99]** with at most two decimals. A pause marker must sit between two pronounceable pieces of text; consecutive markers are not allowed - Paralinguistic tags: only supported when the model is `speech-2.8-hd` or `speech-2.8-turbo`. Supported tags: `(laughs)`, `(chuckle)`, `(coughs)`, `(clear-throat)`, `(groans)`, `(breath)`, `(pant)`, `(inhale)`, `(exhale)`, `(gasps)`, `(sniffs)`, `(sighs)`, `(snorts)`, `(burps)`, `(lip-smacking)`, `(humming)`, `(hissing)`, `(emm)`, `(sneezes)` x-parser-schema-id: x-parser-schema-id: SendTaskContinueEvent title: Task Continue Event description: >- After receiving the `task_started` event from the server, the task officially begins and you can send `task_continue` events carrying the text to synthesize. Key difference from `/ws/v1/t2a_v2`: **text is not synthesized one message at a time**. It enters a server-side sentence buffer and is only sent for synthesis once it forms a sentence, so it is safe to send text character by character (or token by token): - flushed immediately on sentence-ending punctuation (`。！？…；!?.` and newline) - flushed on secondary punctuation (`，、：,;:`) once the buffered segment is long enough - force-split when the buffer reaches its length limit - after you stop sending for a short while, the leftover buffer is flushed automatically Whitespace-only text is discarded silently and returns no error. If no new event is sent within 120s after the last server response, the WebSocket connection closes automatically. example: |- { "event": "task_continue", "text": "Omg(sighs), the real danger is not that computers start thinking like people, but that people start thinking like computers. Computers can only help us with simple tasks." } bindings: [] extensions: - id: x-parser-unique-object-id value: send_task_continue - &ref_7 id: send_task_cancel contentType: application/json payload: - name: Cancel Synthesis Event description: >- Send `task_cancel` to interrupt synthesis immediately: the server discards any text still in the sentence buffer, aborts the in-flight synthesis, and replies with `task_canceled`. Audio already delivered to the client is not revoked. After cancelling, the session returns to the `task_started` state and **you may keep sending `task_continue`** without reconnecting. This is intended for barge-in during real-time conversation. type: object properties: - name: event type: string description: Event type for this step; must be `task_cancel` enumValues: - task_cancel required: true headers: [] jsonPayloadSchema: type: object required: - event properties: event: type: string enum: - task_cancel default: - task_cancel description: Event type for this step; must be `task_cancel` x-parser-schema-id: x-parser-schema-id: SendTaskCancelEvent title: Cancel Synthesis Event description: >- Send `task_cancel` to interrupt synthesis immediately: the server discards any text still in the sentence buffer, aborts the in-flight synthesis, and replies with `task_canceled`. Audio already delivered to the client is not revoked. After cancelling, the session returns to the `task_started` state and **you may keep sending `task_continue`** without reconnecting. This is intended for barge-in during real-time conversation. example: |- { "event": "task_cancel" } bindings: [] extensions: - id: x-parser-unique-object-id value: send_task_cancel - &ref_8 id: send_task_flush contentType: application/json payload: - name: Flush pending text (without ending the session) description: >- Send `task_flush` to make the server synthesize whatever text is still buffered, **without ending the session or closing the connection**. `task_flushed` is returned once that audio has been sent. Use it when a turn is over but its last piece of text is short and unpunctuated. Such a remainder is not flushed by the short silence window (that would risk cutting mid-word), so it otherwise waits for a longer backstop window. `task_flush` gets that audio out immediately. Difference from `task_finish`: `task_finish` closes the connection, so it cannot be used to flush one turn's tail when you run multiple turns over a single connection. After `task_flush` the session stays in `task_started` and you may keep sending `task_continue`. If the buffer happens to be empty, `task_flushed` is still returned normally; that is not an error. type: object properties: - name: event type: string description: The session event type; use `task_flush` for this step enumValues: - task_flush required: true headers: [] jsonPayloadSchema: type: object required: - event properties: event: type: string enum: - task_flush default: - task_flush description: The session event type; use `task_flush` for this step x-parser-schema-id: x-parser-schema-id: SendTaskFlushEvent title: Flush pending text (without ending the session) description: >- Send `task_flush` to make the server synthesize whatever text is still buffered, **without ending the session or closing the connection**. `task_flushed` is returned once that audio has been sent. Use it when a turn is over but its last piece of text is short and unpunctuated. Such a remainder is not flushed by the short silence window (that would risk cutting mid-word), so it otherwise waits for a longer backstop window. `task_flush` gets that audio out immediately. Difference from `task_finish`: `task_finish` closes the connection, so it cannot be used to flush one turn's tail when you run multiple turns over a single connection. After `task_flush` the session stays in `task_started` and you may keep sending `task_continue`. If the buffer happens to be empty, `task_flushed` is still returned normally; that is not an error. example: |- { "event": "task_flush" } bindings: [] extensions: - id: x-parser-unique-object-id value: send_task_flush - &ref_9 id: send_task_finish contentType: application/json payload: - name: Task Finish Event description: >- On receiving `task_finish`, the server first flushes any text still sitting in the sentence buffer, waits until all of that audio has been returned, and only then replies with `task_finished` and closes the connection. As a result **a trailing incomplete sentence is never dropped**, and the client does not need to append punctuation before finishing. type: object properties: - name: event type: string description: >- The type of session event. For this step, set the value to `task_finish`. enumValues: - task_finish required: true headers: [] jsonPayloadSchema: type: object required: - event properties: event: type: string enum: - task_finish default: - task_finish description: >- The type of session event. For this step, set the value to `task_finish`. x-parser-schema-id: x-parser-schema-id: SendTaskFinishEvent title: Task Finish Event description: >- On receiving `task_finish`, the server first flushes any text still sitting in the sentence buffer, waits until all of that audio has been returned, and only then replies with `task_finished` and closes the connection. As a result **a trailing incomplete sentence is never dropped**, and the client does not need to append punctuation before finishing. example: |- { "event": "task_finish" } bindings: [] extensions: - id: x-parser-unique-object-id value: send_task_finish bindings: [] extensions: &ref_2 - id: x-parser-unique-object-id value: t2a_v2_bidi_websocket - &ref_4 id: receiveMessage title: Receive message description: '' type: send messages: - &ref_10 id: receive_connected_success contentType: application/json payload: - name: Connected Success Event description: Notification that T2A task has started type: object properties: - name: session_id type: string description: The ID of the entire session. required: false - name: event type: string description: >- The type of session event. Returns `connected_success` upon successful connection. required: false - name: trace_id type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- Status code, where `0` indicates a successful connection. For more information, please refer to the [Error Code Reference](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false headers: [] jsonPayloadSchema: type: object properties: session_id: type: string description: The ID of the entire session. x-parser-schema-id: event: type: string const: connected_success description: >- The type of session event. Returns `connected_success` upon successful connection. x-parser-schema-id: trace_id: type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. x-parser-schema-id: base_resp: &ref_0 type: object properties: status_code: type: integer description: >- Status code, where `0` indicates a successful connection. For more information, please refer to the [Error Code Reference](/api-reference/errorcode). x-parser-schema-id: status_msg: type: string description: Detailed status message. x-parser-schema-id: x-parser-schema-id: BaseResp connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: x-parser-schema-id: ReceiveConnectedSuccessEvent title: Connected Success Event description: Notification that T2A task has started example: |- { "session_id": "xxxx", "event": "connected_success", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 0, "status_msg": "success" }, "connect_id": "301871346491491" } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_connected_success - &ref_11 id: receive_task_started contentType: application/json payload: - name: Task Started Event description: Notification that T2A task has started type: object properties: - name: session_id type: string description: The id of the entire session. required: false - name: event type: string description: >- The type of session event. For this step, a successful response returns `task_started`. required: false - name: trace_id type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- - 0: success - 2202: illegal event For more information, please refer to the [Error Code Reference](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false headers: [] jsonPayloadSchema: type: object properties: session_id: type: string description: The id of the entire session. x-parser-schema-id: event: type: string const: >- The type of session event. For this step, a successful response returns `task_started`. x-parser-schema-id: trace_id: type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. x-parser-schema-id: base_resp: &ref_1 type: object properties: status_code: type: integer description: >- - 0: success - 2202: illegal event For more information, please refer to the [Error Code Reference](/api-reference/errorcode). x-parser-schema-id: status_msg: type: string description: Detailed status message. x-parser-schema-id: x-parser-schema-id: TaskStartFinishBaseResp connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: x-parser-schema-id: ReceiveTaskStartedEvent title: Task Started Event description: Notification that T2A task has started example: |- { "session_id": "xxxx", "event": "task_started", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 0, "status_msg": "success" }, "connect_id": "301871346491491" } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_task_started - &ref_12 id: receive_sentence_start contentType: application/json payload: - name: Sentence Start Event description: >- Returned when the server has buffered a complete sentence and starts synthesizing it. The audio for that sentence then arrives in chunks via `task_continued` until `sentence_end`. Use it to count how many sentences the server actually grouped the text into type: object properties: - name: session_id type: string description: The id of the entire session. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false - name: event type: string description: >- Event type returned when a buffered sentence starts synthesizing: `sentence_start` required: false - name: trace_id type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- Status code, where `0` indicates a successful connection. For more information, please refer to the [Error Code Reference](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false headers: [] jsonPayloadSchema: type: object properties: session_id: type: string description: The id of the entire session. x-parser-schema-id: connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: event: type: string const: sentence_start description: >- Event type returned when a buffered sentence starts synthesizing: `sentence_start` x-parser-schema-id: trace_id: type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. x-parser-schema-id: base_resp: *ref_0 x-parser-schema-id: ReceiveSentenceStartEvent title: Sentence Start Event description: >- Returned when the server has buffered a complete sentence and starts synthesizing it. The audio for that sentence then arrives in chunks via `task_continued` until `sentence_end`. Use it to count how many sentences the server actually grouped the text into example: |- { "session_id": "my-session-001", "connect_id": "301871346491491", "event": "sentence_start", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 0, "status_msg": "success" } } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_sentence_start - &ref_13 id: receive_task_continued contentType: application/json payload: - name: Task Continued Event description: Notification that T2A task is continuing type: object properties: - name: data type: object description: >- `data` may return `null`. Please check for null when referencing example code. required: false properties: - name: audio type: string description: >- Synthesized audio segment, hex-encoded, generated in the format defined in the request (mp3/ pcm/ flac). required: false - name: trace_id type: string description: >- ID of the individual request within the session, useful for troubleshooting or feedback. required: false - name: session_id type: string description: ID of the entire session. required: false - name: event type: string description: Session event type. On success, returns `"task_continued"`. required: false - name: is_final type: boolean description: Indicates whether this response is the final one. required: false - name: extra_info type: object description: Additional information. required: false properties: - name: audio_length type: integer description: Audio duration in milliseconds required: false - name: audio_sample_rate type: integer description: Audio sample rate required: false - name: audio_size type: integer description: Audio file size in bytes required: false - name: bitrate type: integer description: Audio bitrate required: false - name: audio_format type: string description: 'Generated audio file format. Options: mp3/pcm/flac.' required: false - name: audio_channel type: integer description: 'Number of audio channels: 1 = mono, 2 = stereo.' required: false - name: invisible_character_ratio type: integer description: >- Ratio of illegal characters. If ≤ 10%, audio is generated normally with the ratio returned. Above 10%, an error is triggered. required: false - name: usage_characters type: integer description: Number of billing characters for this speech generation. required: false - name: word_count type: integer description: >- Count of pronounced characters, including Chinese characters, digits, and letters (punctuation excluded). required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- Status code - 0: success - 1000: unknown error - 1001: timeout - 1002: rate limited - 1004: authentication failed - 1039: TPM rate limited - 1042: invalid characters exceed 10% - 2013: invalid input parameters - 2201: connection closed because it was idle for too long (see the Keeping the connection alive section) - 2202: invalid event - 2204: a single `task_continue` text exceeded 10,000 characters and was skipped; the connection and session stay open - 2205: too much text is queued for synthesis (you are sending faster than synthesis can keep up). **This is not a quota rate limit**: the connection and session stay open, so just resend that `task_continue` a moment later. Do not reconnect or send `task_start` again - 2206: illegal event order (for example sending `task_start` again after the task already started) Note: this API discards whitespace-only text silently and never returns `2203`. See the [error code list](/api-reference/errorcode) for details required: false - name: status_msg type: string description: Detailed status message. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false headers: [] jsonPayloadSchema: type: object properties: data: type: object description: >- `data` may return `null`. Please check for null when referencing example code. properties: audio: type: string description: >- Synthesized audio segment, hex-encoded, generated in the format defined in the request (mp3/ pcm/ flac). x-parser-schema-id: x-parser-schema-id: trace_id: type: string description: >- ID of the individual request within the session, useful for troubleshooting or feedback. x-parser-schema-id: session_id: type: string description: ID of the entire session. x-parser-schema-id: event: type: string description: Session event type. On success, returns `"task_continued"`. x-parser-schema-id: is_final: type: boolean description: Indicates whether this response is the final one. x-parser-schema-id: extra_info: type: object description: Additional information. properties: audio_length: type: integer format: int64 description: Audio duration in milliseconds x-parser-schema-id: audio_sample_rate: type: integer format: int64 description: Audio sample rate x-parser-schema-id: audio_size: type: integer format: int64 description: Audio file size in bytes x-parser-schema-id: bitrate: type: integer format: int64 description: Audio bitrate x-parser-schema-id: audio_format: type: string description: 'Generated audio file format. Options: mp3/pcm/flac.' x-parser-schema-id: audio_channel: type: integer format: int64 description: 'Number of audio channels: 1 = mono, 2 = stereo.' x-parser-schema-id: invisible_character_ratio: type: integer format: float description: >- Ratio of illegal characters. If ≤ 10%, audio is generated normally with the ratio returned. Above 10%, an error is triggered. x-parser-schema-id: usage_characters: type: integer format: int64 description: Number of billing characters for this speech generation. x-parser-schema-id: word_count: type: integer format: int64 description: >- Count of pronounced characters, including Chinese characters, digits, and letters (punctuation excluded). x-parser-schema-id: x-parser-schema-id: ExtraInfo base_resp: type: object properties: status_code: type: integer description: >- Status code - 0: success - 1000: unknown error - 1001: timeout - 1002: rate limited - 1004: authentication failed - 1039: TPM rate limited - 1042: invalid characters exceed 10% - 2013: invalid input parameters - 2201: connection closed because it was idle for too long (see the Keeping the connection alive section) - 2202: invalid event - 2204: a single `task_continue` text exceeded 10,000 characters and was skipped; the connection and session stay open - 2205: too much text is queued for synthesis (you are sending faster than synthesis can keep up). **This is not a quota rate limit**: the connection and session stay open, so just resend that `task_continue` a moment later. Do not reconnect or send `task_start` again - 2206: illegal event order (for example sending `task_start` again after the task already started) Note: this API discards whitespace-only text silently and never returns `2203`. See the [error code list](/api-reference/errorcode) for details x-parser-schema-id: status_msg: type: string description: Detailed status message. x-parser-schema-id: x-parser-schema-id: TaskContinueBaseResp connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: x-parser-schema-id: ReceiveTaskContinuedEvent title: Task Continued Event description: Notification that T2A task is continuing example: |- { "data": { "audio": "xxx" }, "extra_info": { "audio_channel": 1, "audio_format": "mp3", "audio_length": 9914, "audio_sample_rate": 32000, "audio_size": 157869, "bitrate": 128000, "invisible_character_ratio": 0, "usage_characters": 158, "word_count": 158 }, "is_final": true, "session_id": "301871346491491", "trace_id": "04ee3794e2c9e4a6d5f99e77742f06fd", "base_resp": { "status_code": 0, "status_msg": "success" }, "connect_id": "301871346491491" } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_task_continued - &ref_14 id: receive_sentence_end contentType: application/json payload: - name: Sentence End Event description: >- Returned when all audio for the current sentence has been delivered. Note the three levels of completion: `is_final` marks the end of audio for one request, `sentence_end` marks the end of the current sentence, and `task_finished` marks the end of the whole session type: object properties: - name: session_id type: string description: The id of the entire session. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false - name: event type: string description: >- Event type returned when a sentence finishes synthesizing: `sentence_end` required: false - name: trace_id type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- Status code, where `0` indicates a successful connection. For more information, please refer to the [Error Code Reference](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false headers: [] jsonPayloadSchema: type: object properties: session_id: type: string description: The id of the entire session. x-parser-schema-id: connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: event: type: string const: sentence_end description: >- Event type returned when a sentence finishes synthesizing: `sentence_end` x-parser-schema-id: trace_id: type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. x-parser-schema-id: base_resp: *ref_0 x-parser-schema-id: ReceiveSentenceEndEvent title: Sentence End Event description: >- Returned when all audio for the current sentence has been delivered. Note the three levels of completion: `is_final` marks the end of audio for one request, `sentence_end` marks the end of the current sentence, and `task_finished` marks the end of the whole session example: |- { "session_id": "my-session-001", "connect_id": "301871346491491", "event": "sentence_end", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 0, "status_msg": "success" } } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_sentence_end - &ref_15 id: receive_task_canceled contentType: application/json payload: - name: Cancelled Event description: >- The server returns `task_canceled` to confirm the interruption took effect. You may continue sending `task_continue` afterwards type: object properties: - name: session_id type: string description: The id of the entire session. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false - name: event type: string description: >- Event type returned once the cancellation takes effect: `task_canceled` required: false - name: trace_id type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- Status code, where `0` indicates a successful connection. For more information, please refer to the [Error Code Reference](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false headers: [] jsonPayloadSchema: type: object properties: session_id: type: string description: The id of the entire session. x-parser-schema-id: connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: event: type: string const: task_canceled description: >- Event type returned once the cancellation takes effect: `task_canceled` x-parser-schema-id: trace_id: type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. x-parser-schema-id: base_resp: *ref_0 x-parser-schema-id: ReceiveTaskCanceledEvent title: Cancelled Event description: >- The server returns `task_canceled` to confirm the interruption took effect. You may continue sending `task_continue` afterwards example: |- { "session_id": "my-session-001", "connect_id": "301871346491491", "event": "task_canceled", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 0, "status_msg": "success" } } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_task_canceled - &ref_16 id: receive_task_flushed contentType: application/json payload: - name: Pending text flushed description: >- The server returns `task_flushed` to indicate that all audio for this `task_flush` has been sent. It is ordered after the audio frames produced by the flush, so receiving it means this turn's audio is complete. You may keep sending `task_continue` afterwards. type: object properties: - name: session_id type: string description: The id of the whole session required: false - name: connect_id type: string description: >- The id of the current WebSocket connection. Unlike `session_id`, there is one `connect_id` per connection while `session_id` identifies the synthesis session inside it required: false - name: event type: string description: >- The session event type; `task_flushed` is returned once the pending text has been flushed required: false - name: trace_id type: string description: >- The id of a single request within the session, useful when reporting issues required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- Status code, where `0` indicates a successful connection. For more information, please refer to the [Error Code Reference](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false headers: [] jsonPayloadSchema: type: object properties: session_id: type: string description: The id of the whole session x-parser-schema-id: connect_id: type: string description: >- The id of the current WebSocket connection. Unlike `session_id`, there is one `connect_id` per connection while `session_id` identifies the synthesis session inside it x-parser-schema-id: event: type: string const: task_flushed description: >- The session event type; `task_flushed` is returned once the pending text has been flushed x-parser-schema-id: trace_id: type: string description: >- The id of a single request within the session, useful when reporting issues x-parser-schema-id: base_resp: *ref_0 x-parser-schema-id: ReceiveTaskFlushedEvent title: Pending text flushed description: >- The server returns `task_flushed` to indicate that all audio for this `task_flush` has been sent. It is ordered after the audio frames produced by the flush, so receiving it means this turn's audio is complete. You may keep sending `task_continue` afterwards. example: |- { "session_id": "my-session-001", "connect_id": "301871346491491", "event": "task_flushed", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 0, "status_msg": "success" } } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_task_flushed - &ref_17 id: receive_task_finished contentType: application/json payload: - name: Task Finished Event description: Notification that T2A task has completed successfully type: object properties: - name: trace_id type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. required: false - name: session_id type: string description: ID of the entire session. required: false - name: event type: string description: Session event type. On success, returns `"task_finished"`. required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- - 0: success - 2202: illegal event For more information, please refer to the [Error Code Reference](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false headers: [] jsonPayloadSchema: type: object properties: trace_id: type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. x-parser-schema-id: session_id: type: string description: ID of the entire session. x-parser-schema-id: event: type: string description: Session event type. On success, returns `"task_finished"`. x-parser-schema-id: base_resp: *ref_1 connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: x-parser-schema-id: ReceiveTaskFinishedEvent title: Task Finished Event description: Notification that T2A task has completed successfully example: |- { "session_id": "xxxx", "event": "task_finished", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 0, "status_msg": "success" }, "connect_id": "301871346491491" } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_task_finished - &ref_18 id: receive_task_failed contentType: application/json payload: - name: Task Failed Event description: >- If the `task_failed` event is received, it indicates that the task has failed. In this case, the WebSocket connection must be closed, and the error should be handled. type: object properties: - name: trace_id type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. required: false - name: session_id type: string description: ID of the entire session. required: false - name: event type: string description: Session event type. On success, returns `"task_finished"`. required: false - name: base_resp type: object required: false properties: - name: status_code type: integer description: >- The status code. - `1000`: Unknown error - `1001`: Timeout - `1002`: Rate limit exceeded - `1004`: Authentication failed - `1039`: TPM rate limit triggered - `1042`: More than 10% invalid characters - `2013`: Invalid input format - `2201`: Connection closed because it was idle for too long (see the Keeping the connection alive section) For more codes, see [Error Codes](/api-reference/errorcode). required: false - name: status_msg type: string description: Detailed status message. required: false - name: connect_id type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection required: false headers: [] jsonPayloadSchema: type: object properties: trace_id: type: string description: >- The ID of a single request within the session, useful for troubleshooting or feedback. x-parser-schema-id: session_id: type: string description: ID of the entire session. x-parser-schema-id: event: type: string description: Session event type. On success, returns `"task_finished"`. x-parser-schema-id: base_resp: type: object properties: status_code: type: integer description: >- The status code. - `1000`: Unknown error - `1001`: Timeout - `1002`: Rate limit exceeded - `1004`: Authentication failed - `1039`: TPM rate limit triggered - `1042`: More than 10% invalid characters - `2013`: Invalid input format - `2201`: Connection closed because it was idle for too long (see the Keeping the connection alive section) For more codes, see [Error Codes](/api-reference/errorcode). x-parser-schema-id: status_msg: type: string description: Detailed status message. x-parser-schema-id: x-parser-schema-id: TaskFailedBaseResp connect_id: type: string description: >- Id of the current WebSocket connection. Distinct from `session_id`: one connection has one `connect_id`, while `session_id` identifies a synthesis session inside that connection x-parser-schema-id: x-parser-schema-id: ReceiveTaskFailedEvent title: Task Failed Event description: >- If the `task_failed` event is received, it indicates that the task has failed. In this case, the WebSocket connection must be closed, and the error should be handled. example: |- { "session_id": "xxxx", "event": "task_failed", "trace_id": "0303a2882bf18235ae7a809ae0f3cca7", "base_resp": { "status_code": 1004, "status_msg": "XXXXXXX" }, "connect_id": "301871346491491" } bindings: [] extensions: - id: x-parser-unique-object-id value: receive_task_failed bindings: [] extensions: *ref_2 sendOperations: - *ref_3 receiveOperations: - *ref_4 sendMessages: - *ref_5 - *ref_6 - *ref_7 - *ref_8 - *ref_9 receiveMessages: - *ref_10 - *ref_11 - *ref_12 - *ref_13 - *ref_14 - *ref_15 - *ref_16 - *ref_17 - *ref_18 extensions: - id: x-parser-unique-object-id value: t2a_v2_bidi_websocket securitySchemes: [] ````