Stream speech

Synthesize text and return the audio as a streaming response.

POST /sound_clone/api/v1/tts/stream

Default model path

curl --location 'https://api.myvocal.ai/sound_clone/api/v1/tts/stream' \
--header 'accessKey: <your_api_key>' \
--header 'Content-Type: application/json' \
--data '{
  "voiceId": "<voice_id>",
  "title": "tts-stream-default",
  "text": "Hello from MyVocal streaming endpoint.",
  "language": "en",
  "voiceSettings": {
    "stability": 0.7
  }
}' \
--output output-default.mp3

200 Success

audio/mpeg stream (binary)

_{Text to Speech · Workflow: TTS quickstart · Models: TTS Models & Migration}

Header

accessKey · string · required

Your API key. See Authentication.

Body

voiceId · string · required

Voice ID returned by GET /sound_clone/api/v1/voices.

title · string

Optional title for the generated record.

text · string · required

Text to synthesize.

language · string

Language selector.

modelId · string

Optional explicit model selector.

voiceSettings · object

For advanced settings, use suggested values unless you have strong tuning needs.

Show Toggle object

stability · number · default: "0.7"

a decimal in the range 0.1 to 1 inclusive, with a recommended default value of 0.7. Values below 0.1 or above 1 are rejected; no fixed step is enforced.

When your parameters are set lower, the emotional variation of the TTS audio will be more abundant. Even for the same text, the effects generated multiple times will be different. You can choose the most suitable one by generating the same text’s TTS audio multiple times.

When your parameters are set higher, the emotion of the TTS audio will tend to be consistent.

Accent model note: a value inside the normal legal range (0.1 to 1) is accepted for compatibility but has no effect on myvocal_v3_accent_enhance. An out-of-range value is still rejected with code = 10006.

For myvocal_v3, the currently publicly documented stability values are:

For myvocal_v3_accent_enhance, plain text is the supported minimal path. The inline [tag] prompt control documented for myvocal_v3 is not documented as equivalent, and the service does not rewrite your text. Send plain text for the Accent model.

Prompt Control (myvocal_v3)

Reference capability source: v3.myvocal.ai. Use [tag] in text to guide style and delivery.

Basic syntax:

[tag] Your text content

Examples:

[happy] Hello everyone, welcome to MyVocal.
[laughter] That was funny.

Representative categories:

• Emotion control: happy, excited, sad, angry, surprised
• Non-speech sounds: laughter, throat clearing, yawning, shushing, screaming
• Sound-related effects: breathing, sighing, gasping, crying, murmuring

Response

Content-Type · string

Success usually returns audio/mpeg.

Body · stream

Binary audio stream payload.

An HTTP 200 does not by itself mean the request succeeded. Read Content-Type before trusting the body:

• audio/* (normally audio/mpeg): the audio payload described above.
• any JSON content type: the standard result envelope with a non-success code (see Common errors below).

The two paths fail differently. Before any audio is sent, a rejected or failed request is answered with the JSON envelope above. After audio has started, the connection is terminated instead of appending a JSON body, because appending JSON to a partial audio body would corrupt it. Treat a short or interrupted download as a failed request, not as a complete audio file.

Default model path

curl --location 'https://api.myvocal.ai/sound_clone/api/v1/tts/stream' \
--header 'accessKey: <your_api_key>' \
--header 'Content-Type: application/json' \
--data '{
  "voiceId": "<voice_id>",
  "title": "tts-stream-default",
  "text": "Hello from MyVocal streaming endpoint.",
  "language": "en",
  "voiceSettings": {
    "stability": 0.7
  }
}' \
--output output-default.mp3

200 Success

audio/mpeg stream (binary)

Common errors

• Invalid language option. (code = 10004): non-V3/default path is missing or using an unsupported language value; or the Accent model received a code outside its own language list.
• VOICE_NOT_FOUND (code = 10005): voiceId is invalid or not available under this API key.
• Invalid stability option. (code = 10006): voiceSettings.stability is outside the legal range.
• code = 10003: the Accent model path was requested but its execution path is unavailable. The request is not silently served by another model.
• code = -1: upstream or backend failed to complete generation.