Try Qwen-Audio 3.0 TTS Flash for low latency text-to-speech. Hear a real multilingual sample, compare voices and controls, and copy API code.
Provider: Alibaba
Category: music generation
Endpoint: /v1/tasks
Status: Available
Cost: See live page pricing
Try Qwen-Audio 3.0 TTS Flash for low latency text-to-speech. Hear a real multilingual sample, compare voices and controls, and copy API code.
Pricing
Standard Usage
Enter the content to convert to speech; up to 20,000 Unicode characters per task; Length: 1 - 20000 characters
Choose a system voice supported by this model
Choose the output audio format; PCM is raw audio data
Output audio sample rate in Hz; Integer only
Output volume from 0 to 100; Range: 0 - 100, Integer only
Speech-rate multiplier; 1 is the normal speed; Range: 0.5 - 2
Pitch multiplier; 1 preserves the original pitch; Range: 0.5 - 2
Applies only when the audio format is Opus, in kbps; Range: 6 - 510, Integer only
Optional hint for the text language; choose at most one; Up to 1 items
Optional control for dialect, emotion, or speaking style; At least 1 characters
Ready for generation
Configure your parameters and click "Run" to see the output here.
Hear a real bilingual output and inspect the text and default voice used for synthesis.
Tests Mandarin clarity, English switching, numbers, rhythm, and natural pauses with the default system voice.
Text input
欢迎来到 HiAPI 音频模型实验室。我们正在比较两种语音合成模型在中文表达、英文切换和数字读法上的表现。Today, we test clarity, rhythm, and natural pauses. 版本三点零，样本编号二零二六。
Choose an adjacent route for music, single-speaker speech, or multi-speaker dialogue.
Multiple speakers
Use separate voices for each line in a structured conversation.
Try Qwen-Audio 3.0 TTS Flash for low latency text-to-speech. Hear a real multilingual sample, compare voices and controls, and copy API code.
The Playground, listening examples, and API Quickstart use the same model ID: qwen-audio-3.0-tts-flash.
These specifications reflect the current Qwen-Audio 3.0 TTS Flash request contract.
Use three steps to organize the input, validate the audio, and integrate the API.
Use punctuation and explicit number formatting to make pauses and pronunciation clear.
Select a system voice, then set rate, pitch, volume, sample rate, and format.
Check pronunciation and rhythm in Playground, then create production tasks with the same model ID.
Copy a minimal request. See Docs for complete parameters, task retrieval, and errors.
POST /v1/tasksqwen-audio-3.0-tts-flashThe create request immediately returns a taskId; audio generation continues asynchronously.
A task accepts 1 to 20,000 Unicode code points. Upstream billing counts CJK ideographs as two effective characters and other characters as one.
Yes. Rate and pitch both range from 0.5 to 2. Output supports PCM, WAV, MP3, and Opus at sample rates from 8 kHz to 48 kHz.
HiAPI bills effective characters and displays the rate per 1,000 characters. See the live page price and task record for the final charge.