alibaba/qwen-3-tts/1.7b
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.
Example output — click Run to generate your own
Send your first request
OpenAI-compatible endpoint with unified authentication and usage tracking.
https://api.sandbase.ai/v1/runalibaba/qwen-3-tts/1.7bInput Schema
15 parameters · 1 required · 14 optional
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt | string | Required | Optional prompt to guide the style of the generated speech. This prompt will be ignored if a speaker embedding is provided. |
text | string | Optional | The text to be converted to speech. |
top_k | integer | Optional | Top-k sampling parameter. · Min: 0 · Default: 50 |
top_p | number | Optional | Top-p sampling parameter. · Min: 0 · Max: 1 · Default: 1 |
voice | string | Optional | 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. · Options: Vivian, Serena, Uncle_Fu, Dylan, Eric, Ryan, Aiden, Ono_Anna, Sohee VivianSerenaUncle_FuDylanEricRyanAidenOno_AnnaSohee |
language | string | Optional | The language of the voice. · Options: Auto, English, Chinese, Spanish, French, German, Italian, Japanese, Korean, Portuguese, Russian · Default: "Auto" AutoEnglishChineseSpanishFrenchGermanItalianJapaneseKoreanPortugueseRussian |
temperature | number | Optional | Sampling temperature; higher => more random. · Min: 0 · Max: 1 · Default: 0.9 |
max_new_tokens | integer | Optional | Maximum number of new codec tokens to generate. · Min: 1 · Max: 8192 · Default: 200 |
reference_text | string | Optional | Optional reference text that was used when creating the speaker embedding. Providing this can improve synthesis quality when using a cloned voice. |
subtalker_top_k | integer | Optional | Top-k for sub-talker sampling. · Min: 0 · Default: 50 |
subtalker_top_p | number | Optional | Top-p for sub-talker sampling. · Min: 0 · Max: 1 · Default: 1 |
repetition_penalty | number | Optional | Penalty to reduce repeated tokens/codes. · Min: 0 · Default: 1.05 |
subtalker_dosample | boolean | Optional | Sampling switch for the sub-talker. · Default: true |
subtalker_temperature | number | Optional | Temperature for sub-talker sampling. · Min: 0 · Max: 1 · Default: 0.9 |
speaker_voice_embedding_file_url | string | Optional | 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. |
Output Schema
| Field | Type | Description |
|---|---|---|
id | string | Unique identifier for the generation task |
status | string | Task status: pending, running, completed, failed, timeout |
model | string | Model used for the generation |
outputs | array | Array of output items |
outputs[].url | string | URL of the generated artifact |
outputs[].content_type | string | MIME type (e.g. image/png, video/mp4) |
error | object | null | Error details if failed, null on success |
error.type | string | Machine-readable error type code |
error.message | string | Human-readable error description |
Async Workflow
This model uses asynchronous execution. Submit a request and poll for the result.
- Submit — POST to /v1/run, receive an
id - Poll — GET /v1/run/{id} until status is
completed,failed, ortimeout - Retrieve — Read
outputsfrom the completed response
Code Examples
Ready-to-run snippets
# Step 1: Submit
curl -X POST https://api.sandbase.ai/v1/run \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "alibaba/qwen-3-tts/1.7b",
"text": "I am solving the equation: x = [-b ± √(b²-4ac)] / 2a? Nobody can — it's a disaster (◍•͈⌔•͈◍), very sad!",
"top_k": 50,
"top_p": 1,
"voice": "Vivian",
"prompt": "Very happy.",
"language": "English",
"temperature": 0.9,
"max_new_tokens": 200,
"reference_text": "Okay. Yeah. I resent you. I love you. I respect you. But you know what? You blew it! And it is all thanks to you.",
"subtalker_top_k": 50,
"subtalker_top_p": 1,
"repetition_penalty": 1.05,
"subtalker_dosample": true,
"subtalker_temperature": 0.9,
"speaker_voice_embedding_file_url": "https://static.sandbase.ai/examples/alibaba/qwen-3-tts/1.7b/output_url_0.safetensors"
}'
# Step 2: Poll result (replace <id>)
curl https://api.sandbase.ai/v1/run/<id> \
-H "Authorization: Bearer YOUR_API_KEY"API README
Qwen 3 TTS 1.7B
Qwen 3 TTS 1.7B generates expressive speech from text while using natural-language instructions to shape tone, emotion, pacing, and vocal style. The larger route is intended for voice work where nuanced interpretation, stable speaker character, and convincing prosody matter across more demanding scripts.
The model supports multilingual and cross-lingual synthesis, including code-switched passages, and can work with designed voices or a compatible cloned-speaker representation. These abilities make it useful for premium narration, localization, character dialogue, conversational experiences, and emotionally directed voiceover.
Highlights
Nuanced instruction following. Interprets detailed direction for emotional state, delivery style, pace, and vocal attitude.
Rich expressive prosody. Produces natural rhythm, emphasis, pauses, and intonation across demanding scripts.
Multilingual and code-switched speech. Synthesizes multiple languages and mixed-language content without changing workflows.
Stable speaker character. Maintains vocal identity across designed voices and compatible cloned-speaker inputs while changing expression.
Pricing
| Text allowance | Price |
|---|---|
| Each started block of 1,000 characters | $0.090 |
When to Use
| ✅ Good fit | ❌ Consider alternatives |
|---|---|
| The model's named workflow matches the source material and intended output | A different input modality or model route is required |
| A managed asynchronous result is suitable for the production pipeline | A synchronous, interactive editor is essential |
| The documented controls cover the required duration, framing, or format | The project needs controls outside this endpoint's schema |
| Creative iteration benefits from a repeatable request structure | Exact deterministic pixels, frames, geometry, or samples are mandatory |
| A finished downloadable media asset is the desired deliverable | Editable source layers or a native project file are required |
Prompt Guide
For speech synthesis, state the intended result first, then add the subject or source treatment, progression, style, and delivery constraints. Keep one creative variable per phrase, use the documented field names for controls, and change one setting at a time when comparing results.
{
"prompt": "Very happy.",
"text": "I am solving the equation: x = [-b ± √(b²-4ac)] / 2a? Nobody can — it's a disaster (◍•͈⌔•͈◍), very sad!"
}
Technical Specs
| Spec | Value |
|---|---|
| Model ID | alibaba/qwen-3-tts/1.7b |
| Inputs | language, max_new_tokens, prompt, reference_text, repetition_penalty, speaker_voice_embedding_file_url, subtalker_dosample, subtalker_temperature, subtalker_top_k, subtalker_top_p, temperature, text, top_k, top_p, voice |
| Required inputs | prompt |
| Output fields | content_type, url |
| Execution | Async (submit, then poll for result) |
| Language | Auto / English / Chinese / Spanish / French / German / Italian / Japanese / Korean / Portuguese / Russian |
Related Models
alibaba/qwen-3-tts/0.6b— Compare a nearby route in the same local model family.alibaba/qwen-3-tts/clone-voice/0.6b— Compare a nearby route in the same local model family.alibaba/qwen-3-tts/clone-voice/1.7b— Compare a nearby route in the same local model family.