Convert complete text to speech over HTTP. Explore model features, make your first request, and improve synthesis quality.
Features
Non-real-time speech synthesis accepts complete text in one HTTP request and supports non-streaming and SSE streaming output. Non-streaming output returns the complete audio after synthesis finishes. Streaming output returns audio in chunks.
- Languages and dialects: Select a model and voice for the target language, or use an instruction to specify a dialect. See Dialects.
- Expressive speech: Adjust delivery with natural-language instructions, or insert emotion and nonverbal tags into the text. See Instruction control and Emotion and nonverbal tags.
- Custom voices: Create a voice from an audio sample with voice cloning, or from a text description with voice design.
- Fine-grained control and formulas: Use SSML and LaTeX to control pauses and pronunciation, or convert formulas to speech.
- Supported features vary by model and voice. Check the scope in the corresponding feature documentation.
- To submit text incrementally, use real-time speech synthesis.
Get started
- Obtain an API key and set the environment variable
DASHSCOPE_API_KEY. Use a key for the region in the example. - The Qwen-Audio-TTS examples use the Singapore endpoint. For another region, replace
ap-southeast-1in the URL. The CosyVoice examples use China North 2 (Beijing), the only region where CosyVoice is available through this API. See Regions. - Install the example dependency:
- Qwen-Audio-TTS (Recommended)
- CosyVoice
The following examples use The audio is saved as
qwen-audio-3.1-tts-flash and the longanhuan_v3.1 system voice in Singapore to demonstrate non-streaming and streaming output.- Non-streaming output
- Streaming output
output.wav in the current directory.Improve synthesis quality
Languages and pronunciation
If numbers, abbreviations, or symbols are pronounced incorrectly, use language_hints to specify the target synthesis language, or use hot_fix to customize pronunciation and text replacement. For supported models and configuration, see the Qwen-Audio-TTS API or CosyVoice API.
Use SSML to control pauses, the reading of numbers, or word pronunciation. For supported Qwen-Audio-TTS and CosyVoice models, voices, interfaces, and tags, see SSML and LaTeX.
Instruction control
Add instruction to the request’s input object to specify the synthesis language and describe the speech rate, emotion, or speaking style in natural language.
To create a new voice with a specific gender, age, or vocal character, or for guidance and examples on writing voice descriptions, see voice design.
- Qwen-Audio-TTS
- CosyVoice
System, cloned, and designed voices for
qwen-audio-3.1-tts-plus, qwen-audio-3.1-tts-flash, qwen-audio-3.0-tts-plus, and qwen-audio-3.0-tts-flash all support natural-language instructions through instruction in synthesis requests.We recommend specifying the synthesis language and desired effect in instruction to improve synthesis quality. The language specified in the instruction must match the language of the text to be synthesized. For example, when synthesizing Dutch text, set:Dialects
Use dialect synthesis to generate speech in Chinese dialects such as Henan or Sichuan dialect. How you select a dialect depends on the model and voice type.
- Qwen-Audio-TTS
- CosyVoice
- System voices: Select a voice from the Qwen-Audio-TTS voice list that supports the dialect. A voice that directly supports the dialect requires no additional setting. For voices that support dialect instructions, use
instruction. - Cloned voices: Specify a dialect with
instruction, such as请用河南话表达。. - Designed voices: Dialects are not supported.
instruction="请讲上海话。" in the request's input object to synthesize Chinese text in Shanghainese.Emotion and nonverbal tags
Applies to
qwen-audio-3.1-tts-plus, qwen-audio-3.1-tts-flash, qwen-audio-3.0-tts-plus, and qwen-audio-3.0-tts-flash.text, without additional parameters.
Usage examples
This text starts with an excited delivery, inserts laughter, and then continues speaking:
Supported tags
- Control tags
- Nonverbal tags
A control tag sets the emotion or style of the following text until the next control tag. When long sentences are automatically segmented, the effect is bounded by the resulting text segment.
| Tag | Description |
|---|---|
[sad] | Sad |
[amazed] | Amazed |
[deep and loud shouting] | Deep, loud shouting |
[trembling] | Trembling |
[angry] | Angry |
[excited] | Excited |
[sarcastic] | Sarcastic |
[curious] | Curious |
[like dracula] | Dracula-like delivery |
[bored] | Bored |
[tired] | Tired |
[scornful] | Scornful |
[shouting] | Shouting |
[asmr] | Soft ASMR delivery |
[panicked] | Panicked |
[mischievously] | Mischievous |
[empathetic] | Empathetic |
[whispers] | Whispering |
[reluctantly] | Reluctant |
[crying] | Crying |
[serious] | Serious |
[very slowly] | Very slow |
[very fast] | Very fast |
Apply in production
For audio downloads and storage, format and player compatibility, concurrency management, and troubleshooting, see Speech synthesis best practices. Keep request frequency and concurrency within the model's rate limits.
Supported models
- Singapore
- China North 2 (Beijing)
- Qwen-Audio-TTS:
qwen-audio-3.1-tts-plus,qwen-audio-3.1-tts-flash,qwen-audio-3.0-tts-plus,qwen-audio-3.0-tts-flash - Qwen3-TTS:
qwen3-tts-flash