> ## Documentation Index > Fetch the complete documentation index at: https://docs.powertokens.ai/llms.txt > Use this file to discover all available pages before exploring further. # Synchronous TTS HTTP ## OpenAPI ````yaml api-reference/en/zmodelAudio/minimax/audio-generation/t2a-v2.json POST /minimax/v1/t2a_v2 openapi: 3.0.1 info: title: MiniMax T2A API description: Use this API for synchronous text-to-speech synthesis over HTTP. license: name: MIT version: 1.0.0 servers: - url: https://api.powertokens.ai security: - bearerAuth: [] paths: /minimax/v1/t2a_v2: post: tags: - Text to Audio summary: Text to Audio V2 operationId: t2aV2 parameters: - name: Content-Type in: header required: true description: >- Media type of the request body. Set to `application/json` to ensure the request data is in JSON format. schema: type: string enum: - application/json default: application/json requestBody: description: '' content: application/json: schema: $ref: '#/components/schemas/T2aV2Req' examples: Non-streaming: value: model: speech-2.8-hd text: Today is such a happy day (laughs), of course! stream: false voice_setting: voice_id: male-qn-qingse speed: 1 vol: 1 pitch: 0 emotion: happy pronunciation_dict: tone: - 处理/(chu3)(li3) - 危险/dangerous audio_setting: sample_rate: 32000 bitrate: 128000 format: mp3 channel: 1 subtitle_enable: false Streaming: value: model: speech-2.8-hd text: Today is such a happy day (laughs), of course! stream: true voice_setting: voice_id: male-qn-qingse speed: 1 vol: 1 pitch: 0 emotion: happy pronunciation_dict: tone: - 处理/(chu3)(li3) - 危险/dangerous audio_setting: sample_rate: 32000 bitrate: 128000 format: mp3 channel: 1 subtitle_enable: false required: true responses: '200': description: '' content: application/json: schema: $ref: '#/components/schemas/T2aV2Resp' examples: Non-streaming: value: data: audio: status: 2 extra_info: audio_length: 9900 audio_sample_rate: 32000 audio_size: 160323 bitrate: 128000 word_count: 52 invisible_character_ratio: 0 usage_characters: 26 audio_format: mp3 audio_channel: 1 trace_id: 01b8bf9bb7433cc75c18eee6cfa8fe21 base_resp: status_code: 0 status_msg: success Streaming: value: - data: audio: hex-encoded audio_chunk1 status: 1 trace_id: 01b8bf9bb7433cc75c18eee6cfa8fe21 base_resp: status_code: 0 status_msg: '' - data: audio: hex-encoded audio_chunk2 status: 1 trace_id: 01b8bf9bb7433cc75c18eee6cfa8fe21 base_resp: status_code: 0 status_msg: '' - data: audio: hex-encoded audio status: 2 extra_info: audio_length: 6931 audio_sample_rate: 32000 audio_size: 111789 bitrate: 128000 word_count: 112 invisible_character_ratio: 0 usage_characters: 112 audio_format: mp3 audio_channel: 1 trace_id: 04ece790375f3ca2edbb44e8c4c200bf base_resp: status_code: 0 status_msg: success text/event-stream: schema: $ref: '#/components/schemas/T2aV2Resp' examples: Streaming: value: - data: audio: hex-encoded audio_chunk1 status: 1 trace_id: 01b8bf9bb7433cc75c18eee6cfa8fe21 base_resp: status_code: 0 status_msg: '' - data: audio: hex-encoded audio_chunk2 status: 1 trace_id: 01b8bf9bb7433cc75c18eee6cfa8fe21 base_resp: status_code: 0 status_msg: '' - data: audio: hex-encoded audio status: 2 extra_info: audio_length: 6931 audio_sample_rate: 32000 audio_size: 111789 bitrate: 128000 word_count: 112 invisible_character_ratio: 0 usage_characters: 112 audio_format: mp3 audio_channel: 1 trace_id: 04ece790375f3ca2edbb44e8c4c200bf base_resp: status_code: 0 status_msg: success components: schemas: T2aV2Req: type: object required: - model - text properties: model: type: string description: >- Model version. Available options: `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo`. enum: - speech-2.8-hd - speech-2.8-turbo - speech-2.6-hd - speech-2.6-turbo - speech-02-hd - speech-02-turbo text: type: string description: >- Text to be synthesized into speech. Maximum length is less than 10000 characters. If the text exceeds 3000 characters, streaming output is recommended. - Use newline characters to mark paragraph breaks. - Pause control: Supports custom speech time intervals between text segments for customized pause effects. Usage: Insert `<#x#>` markers in the text, where `x` is the pause duration in seconds, range [0.01, 99.99], with up to two decimal places. Pause markers must be placed between two pronounceable text segments; consecutive pause markers are not supported. - Inline pronunciation replacement: Wrap Mandarin Pinyin (with tone numbers `1–5`), IPA symbols, or Cantonese Pinyin (with tone numbers `1–6`) in English parentheses to temporarily override the pronunciation of problematic words or polyphonic characters. - `"The word live is pronounced (lɪv) as a verb and (laɪv) as an adjective."` - `"This is (he2)平, not (huo4)面."` - `"去街市買啲(sung3)。"` - Emotion tag: Only when model is `speech-2.8-hd` or `speech-2.8-turbo`, emotion tags can be inserted in the text. Supported tags: `(laughs)`, `(chuckle)`, `(coughs)`, `(clear-throat)`, `(groans)`, `(breath)`, `(pant)`, `(inhale)`, `(exhale)`, `(gasps)`, `(sniffs)`, `(sighs)`, `(snorts)`, `(burps)`, `(lip-smacking)`, `(humming)`, `(hissing)`, `(emm)`, `(sneezes)`. stream: type: boolean description: >- Controls whether to use streaming output. Defaults to false (non-streaming). stream_options: $ref: '#/components/schemas/T2AStreamOption' voice_setting: $ref: '#/components/schemas/T2AVoiceSetting' audio_setting: $ref: '#/components/schemas/T2AAudioSetting' pronunciation_dict: $ref: '#/components/schemas/PronunciationDict' timbre_weights: type: array items: $ref: '#/components/schemas/TimbreWeights' language_boost: type: string description: >- Whether to enhance recognition for specific minority languages and dialects. Defaults to `null`. Can be set to `auto` for the model to determine automatically. Note: speech-02 series models do not support Persian, Filipino, and 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 default: null voice_modify: $ref: '#/components/schemas/VoiceModify' subtitle_enable: type: boolean description: >- Controls whether to enable subtitle service. Defaults to false. Only effective for `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo` models. default: false subtitle_type: type: string description: >- Subtitle granularity. Defaults to `sentence`. Available values: - `sentence`: Sentence-level timestamps - `word`: Word-level timestamps - `word_streaming`: Streaming-optimized word-level timestamps, only effective when `stream=true`. enum: - sentence - word - word_streaming default: sentence output_format: type: string description: >- Controls the output format. Available values: [`url`, `hex`]. Defaults to `hex`. This parameter only takes effect in non-streaming mode; streaming mode only returns hex format. Returned URLs are valid for 24 hours. enum: - url - hex default: hex aigc_watermark: type: boolean description: >- Controls whether to add an audio rhythm identifier at the end of the synthesized audio. Defaults to false. This parameter only takes effect for non-streaming synthesis. default: false T2aV2Resp: type: object properties: data: type: object description: >- Synthesized audio data object. May be null, so null-checking is required. properties: audio: type: string description: >- Synthesized audio data in hex encoding, matching the output format specified in the request. subtitle_file: type: string description: >- Download link for the synthesized subtitle file. The subtitle file corresponding to the audio, accurate to sentence level (up to 50 characters), in milliseconds, in JSON format. status: type: integer description: >- Current audio stream status: 1 indicates synthesizing, 2 indicates synthesis complete. trace_id: type: string description: >- Session ID for this request, used to help locate issues when consulting or providing feedback. extra_info: type: object description: Additional audio information. properties: audio_length: type: integer format: int64 description: Audio duration in milliseconds. audio_sample_rate: type: integer format: int64 description: Audio sample rate. audio_size: type: integer format: int64 description: Audio file size in bytes. bitrate: type: integer format: int64 description: Audio bitrate. audio_format: type: string description: >- Format of the generated audio file. Available values: `[mp3, pcm, flac]`. enum: - mp3 - pcm - flac audio_channel: type: integer format: int64 description: 'Number of audio channels. 1: mono, 2: stereo.' invisible_character_ratio: type: number format: double description: >- Ratio of invalid characters. If invalid characters do not exceed 10% (inclusive), audio will be generated normally and the invalid character ratio data will be returned; if it exceeds 10%, an error will be reported. usage_characters: type: integer format: int64 description: Number of billable characters. word_count: type: integer format: int64 description: >- Count of pronounced characters, including Chinese characters, digits, and letters, excluding punctuation marks. base_resp: type: object description: Status code and details for this request. properties: status_code: type: integer format: int64 description: >+ Status codes: You can retrieve the trace_id from the response header to help locate issues when consulting or providing feedback. - `0`: Request successful - `1000`: Unknown error - `1001`: Timeout - `1002`: Rate limit triggered - `1004`: Authentication failed - `1039`: TPM rate limit triggered - `1042`: Invalid characters exceed 10% - `2013`: Abnormal input parameters status_msg: type: string description: Status details. T2AStreamOption: type: object properties: exclude_aggregated_audio: type: boolean description: >- Sets whether the last chunk contains the concatenated full audio hex data. Defaults to false, meaning the last chunk includes the concatenated complete audio hex data. T2AVoiceSetting: type: object required: - voice_id properties: voice_id: type: string description: "Voice ID for the synthesized audio. If you need to set mixed voices, use the `timbre_weights` parameter and set this parameter to empty. Supports system voices, cloned voices, and text-to-voice types. Below are some of the latest system voices (ID):\n\n - **Chinese**:\n\t- moss_audio_ce44fc67-7ce3-11f0-8de5-96e35d26fb85\n\t- moss_audio_aaa1346a-7ce7-11f0-8e61-2e6e3c7ee85d\n\t- Chinese (Mandarin)_Lyrical_Voice\n\t- Chinese (Mandarin)_HK_Flight_Attendant\n- **English**:\n\t- English_Graceful_Lady\n\t- English_Insightful_Speaker\n\t- English_radiant_girl\n\t- English_Persuasive_Man\n\t- moss_audio_6dc281eb-713c-11f0-a447-9613c873494c\n\t- moss_audio_570551b1-735c-11f0-b236-0adeeecad052\n\t- moss_audio_ad5baf92-735f-11f0-8263-fe5a2fe98ec8\n\t- English_Lucky_Robot\n- **Japanese**:\n\t- Japanese_Whisper_Belle\n\t- moss_audio_24875c4a-7be4-11f0-9359-4e72c55db738\n\t- moss_audio_7f4ee608-78ea-11f0-bb73-1e2a4cfcd245\n\t- moss_audio_c1a6a3ac-7be6-11f0-8e8e-36b92fbb4f95\n- **Cantonese** (requires `language_boost` set to `Chinese,Yue`):\n\t- Cantonese_GentleLady\n\t- Cantonese_podacast_host_1" speed: type: number format: float description: >- Speech speed of the synthesized audio. Higher values result in faster speech. Value range `[0.5, 2]`. Defaults to 1.0. minimum: 0.5 maximum: 2 default: 1 vol: type: number format: float description: >- Volume of the synthesized audio. Higher values result in louder audio. Value range `(0, 10]`. Defaults to 1.0. minimum: 0 exclusiveMinimum: true maximum: 10 default: 1 pitch: type: integer description: >- Pitch of the synthesized audio. Value range `[-12, 12]`. Defaults to 0, which outputs the original voice pitch. minimum: -12 maximum: 12 default: 0 emotion: type: string description: "Controls the emotion of the synthesized speech. Parameter range: `[\"happy\", \"sad\", \"angry\", \"fearful\", \"disgusted\", \"surprised\", \"calm\", \"fluent\", \"whisper\"]`, corresponding to 9 emotions: happy, sad, angry, fearful, disgusted, surprised, calm, vivid, whisper.\r\n- The model automatically matches the appropriate emotion based on the input text; manual specification is generally not needed.\r\n- This parameter only takes effect for `speech-2.8-hd`, `speech-2.8-turbo`, `speech-2.6-hd`, `speech-2.6-turbo`, `speech-02-hd`, `speech-02-turbo` models.\r\n- Options `fluent` and `whisper` only take effect for `speech-2.6-turbo` and `speech-2.6-hd` models. `speech-2.8-hd` and `speech-2.8-turbo` models do not support `whisper`." enum: - happy - sad - angry - fearful - disgusted - surprised - calm - fluent - whisper text_normalization: type: boolean description: >- Whether to enable Chinese and English text normalization. When enabled, it improves performance in numeric reading scenarios but slightly increases latency. Defaults to false. default: false latex_read: type: boolean description: >- Controls whether to read LaTeX formulas aloud. Defaults to false. **Note**: - Only supports Chinese. When this parameter is enabled, `language_boost` will be set to `Chinese`. - Formulas in the request must be wrapped with `$$` at the beginning and end. - If a formula in the request contains `"\"`, it must be escaped as `"\\"`. Example: The basic quadratic formula ![The quadratic formula](https://filecdn.minimax.chat/public/d6f62e9a-cd3f-4f55-a237-257eef531683.png) Should be represented as `$$x = \\frac{-b \\pm \\sqrt{b^2 - 4ac}}{2a}$$`. default: false T2AAudioSetting: type: object properties: sample_rate: type: integer format: int64 description: >- Sample rate of the generated audio. Available values: `[8000, 16000, 22050, 24000, 32000, 44100]`. Defaults to `32000`. bitrate: type: integer format: int64 description: >- Bitrate of the generated audio. Available values: `[32000, 64000, 128000, 256000]`. Defaults to `128000`. This parameter only takes effect for `mp3` format audio. format: type: string description: >- Format of the generated audio. `pcmu_raw` and `pcmu_wav` are G.711 μ-law encoding (sample rate 8 kHz; `pcmu_raw` is raw data without header, `pcmu_wav` is wrapped in a WAV container). `opus` is Ogg/Opus encoding. enum: - mp3 - pcm - flac - wav - pcmu_raw - pcmu_wav - opus default: mp3 channel: type: integer format: int64 description: >- Number of audio channels. Available values: `[1, 2]`, where `1` is mono and `2` is stereo. Defaults to 1. force_cbr: type: boolean description: >- Constant bitrate (CBR) control for audio. Available values: `false`, `true`. When set to `true`, audio encoding will use constant bitrate mode. Note: This parameter only takes effect when audio is set to **streaming** output and the audio format is `mp3`. default: false PronunciationDict: type: object properties: tone: type: array description: >- Defines custom pronunciation rules in the format `original/replacement`. The replacement can be plain text, parenthesized Pinyin (with tone numbers), parenthesized IPA symbols, or Japanese kana (hiragana, katakana, romaji — can be mixed). Multiple rules take effect simultaneously. Example: ```json dark "pronunciation_dict": { "tone": [ "resume/(rɪˈzjuːm)", "read/(riːd)", "处理/(chu3)(li3)", "郑栅洁/(zheng4)(shan1)杰", "20日/はつか", "4桁/よんけた", "東京/トウキョウ", "omg/oh my god", "の/no" ] } ``` This example demonstrates: IPA transcription for English polyphones, Pinyin + tone numbers for Chinese, mixed Pinyin and Chinese characters, Japanese kana for counters and kanji readings, text-to-text expansion, and symbol replacement. Chinese tones: `1` = first tone, `2` = second tone, `3` = third tone, `4` = fourth tone, `5` = neutral tone. items: type: string TimbreWeights: type: object required: - voice_id - weight properties: voice_id: type: string description: >- Voice ID for the synthesized audio. Must be filled together with the `weight` parameter. Supports system voices, cloned voices, and text-to-voice types. weight: type: integer format: int64 description: >- Weight of each voice in the synthesized audio. Must be filled together with `voice_id`. Value range [1, 100]. Supports up to 4 voice mixtures. The higher the proportion of a single voice, the closer the synthesized voice resembles that voice. ```json dark "timbre_weights": [ { "voice_id": "female-chengshu", "weight": 30 }, { "voice_id": "female-tianmei", "weight": 70 } ] ``` minimum: 1 maximum: 100 VoiceModify: type: object description: |- Voice effect settings. Supported audio formats: - Non-streaming: `mp3`, `wav`, `flac` - Streaming: `mp3` properties: pitch: type: integer description: >- Pitch adjustment (deeper/brighter). Range [-100, 100]. Values closer to -100 make the voice deeper; closer to 100 make it brighter. ![pitch adjustment](https://filecdn.minimax.chat/public/5d210c47-4236-4e81-893b-16cc1ef0302d.png) minimum: -100 maximum: 100 intensity: type: integer description: >- Intensity adjustment (powerful/soft). Range [-100, 100]. Values closer to -100 make the voice more powerful; closer to 100 make it softer. ![intensity adjustment](https://filecdn.minimax.chat/public/862d493e-71d5-4d1f-b7c3-9ac51890631b.png) minimum: -100 maximum: 100 timbre: type: integer description: >- Timbre adjustment (magnetic/crisp). Range [-100, 100]. Values closer to -100 make the voice fuller; closer to 100 make it crisper. ![timbre adjustment](https://filecdn.minimax.chat/public/5f0e6cae-363a-452b-8d42-fbc4ef5a0510.png) minimum: -100 maximum: 100 sound_effects: type: string description: >- Sound effect setting. Only one can be selected at a time. Available values: 1. spacious_echo 2. auditorium_echo 3. lofi_telephone 4. robotic enum: - spacious_echo - auditorium_echo - lofi_telephone - robotic securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: JWT description: 'Pass `Authorization: Bearer ` in the request header.' ````