Qwen 3 TTS - Text to Speech [1.7B]

POST/v1/run

Qwen 3 Tts 1.7b is Alibaba's text-to-speech AI model. Generate human-like voiceovers with expressive intonation, multilingual support, and customizable voice characteristics.

Request body

Submit an async generation request. The model field selects the model; other fields are model-specific input parameters.

stringmodelrequired

Model identifier. Set to alibaba/qwen-3-tts/1.7b.

Default: alibaba/qwen-3-tts/1.7b

stringpromptrequired

Optional prompt to guide the style of the generated speech. This prompt will be ignored if a speaker embedding is provided.

Optional<string>text

The text to be converted to speech.

Optional<number>subtalker_temperature

Temperature for sub-talker sampling.

Range: 0 to 1

Default: 0.9

Optional<string>speaker_voice_embedding_file_url

URL to a speaker embedding file in safetensors format, from `fal-ai/qwen-3-tts/clone-voice` endpoint. If provided, the TTS model will use the cloned voice for synthesis instead of the predefined voices.

Optional<number>repetition_penalty

Penalty to reduce repeated tokens/codes.

Range: 0 to ∞

Default: 1.05

Optional<string>language

The language of the voice.

Allowed values: Auto, English, Chinese, Spanish, French, German, Italian, Japanese, Korean, Portuguese, Russian

Default: Auto

Optional<number>subtalker_top_p

Top-p for sub-talker sampling.

Range: 0 to 1

Default: 1

Optional<number>top_p

Top-p sampling parameter.

Range: 0 to 1

Default: 1

Optional<string>reference_text

Optional reference text that was used when creating the speaker embedding. Providing this can improve synthesis quality when using a cloned voice.

Optional<integer>top_k

Top-k sampling parameter.

Range: 0 to ∞

Default: 50

Optional<integer>subtalker_top_k

Top-k for sub-talker sampling.

Range: 0 to ∞

Default: 50

Optional<number>temperature

Sampling temperature; higher => more random.

Range: 0 to 1

Default: 0.9

Optional<string>voice

The voice to be used for speech synthesis, will be ignored if a speaker embedding is provided. Check out the **[documentation](https://github.com/QwenLM/Qwen3-TTS/tree/main?tab=readme-ov-file#custom-voice-generate)** for each voice's details and which language they primarily support.

Allowed values: Vivian, Serena, Uncle_Fu, Dylan, Eric, Ryan, Aiden, Ono_Anna, Sohee

Optional<integer>max_new_tokens

Maximum number of new codec tokens to generate.

Range: 1 to 8192

Default: 200

Optional<boolean>subtalker_dosample

Sampling switch for the sub-talker.

Default: true

Response Schema

The submit endpoint returns an accepted generation task. Poll the result endpoint with the returned id for terminal outputs or errors.

Optional<string>error

Error message if the task failed. Empty on success.

stringidrequired

Unique identifier for the generation task.

Optional<string>model

Model ID used for the prediction.

Optional<array>outputs

Array of generated content. Empty when status is not completed.

stringstatusrequired

Status of the task: pending, running, completed, failed, or timeout.

Allowed values: pending, running, completed, failed, timeout

Model capabilities

array<string>capability_tagsrequired

Capabilities declared by the model registry.

Default: text-to-speech

stringexecution_moderequired

Execution mode declared by the model registry.

Default: async

# 1. Submit a generation request
curl -X POST https://api.sandbase.ai/v1/run \
  -H "Authorization: Bearer $SANDBASE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "alibaba/qwen-3-tts/1.7b",
  "prompt": "Very happy.",
  "text": "I am solving the equation: x = [-b ± √(b²-4ac)] / 2a? Nobody can — it's a disaster (◍•͈⌔•͈◍), very sad!",
  "voice": "Vivian"
}'

# 2. If the response is still running, poll the returned id
curl https://api.sandbase.ai/v1/run/run_abc123 \
  -H "Authorization: Bearer $SANDBASE_API_KEY"

Response202 Accepted

{
  "id": "run_abc123",
  "model": "alibaba/qwen-3-tts/1.7b",
  "status": "running",
  "outputs": []
}