01
What TTS-Next is designed to generate
qwen-audio-3.1-tts-next is more than a conventional text-to-speech endpoint. Alibaba documents unified generation of speech, multi-speaker dialogue, sound effects and ambient audio, making it relevant to podcasts, short-form video, games and narrative audio.
The model is served through Alibaba Cloud Model Studio and returns completed audio rather than a continuous low-latency voice stream. Use Qwen-Audio Realtime instead when the user needs an interactive spoken assistant.
02
Reference audio and prompt limits
The API accepts a text prompt and up to three optional reference clips. Alibaba currently documents a maximum of 30 seconds and 10 MB per reference, with WAV, MP3 and OGG Opus accepted for references.
The text prompt is capped at 3,000 characters. Podcast scenarios can generate up to 240 seconds in one request, while other documented scenarios are capped at 120 seconds, so longer productions should be segmented deliberately.
03
A practical generation workflow
Start with the core dialogue, then add speaker references, delivery instructions and environmental sound only where needed. Keep reference naming consistent so multi-speaker prompts remain interpretable.
The API returns an output audio URL whose documented lifetime is temporary. Production applications should copy completed assets to durable storage rather than treating the result URL as permanent hosting.
04
Pricing, rate limits and output formats
For China Beijing, Alibaba currently lists $0.848 per million input tokens and $1.696 per million output tokens and notes that list pricing excludes promotions. The documented rate limit is three requests per second.
WAV, MP3 and PCM are documented output formats. Pricing and availability are region-sensitive, so recheck the live model page instead of applying a launch headline or another Qwen audio model's price to TTS-Next.
05
Reference voices, rights and production caveats
Reference-audio support is a technical capability, not permission to imitate a person. Use reference voices only when the required rights and consent are in place, and preserve those governance decisions in production workflows.
TTS-Next currently documents Chinese and English, is non-streaming, and produces a mixed audio asset rather than separate editable stems. Validate consistency, editability and quality on the exact kind of content you plan to publish.
Sources
Primary and supporting sources
Use the linked documentation to check current pricing, access and product behaviour. The dates above describe this article; they do not guarantee that every vendor claim remains current.