Qwen3-TTS Lets You Steer Delivery via Instructions Without Changing the Voice
Delivery control via a text instruction means one voice asset can cover an entire emotional range, cutting voice production costs for dialogue-heavy apps. It also simplifies pipelines: no SSML markup or multi-voice libraries, just prompt text alongside the script. Developers building voice agents, audiobook or dubbing tools, and teams localising content for ASEAN markets can test different instruction phrasings to lock in consistent delivery across languages.
A demonstration from Together shows that Qwen3-TTS can change how a voice delivers a line while keeping the same voice identity. In the example, a voice named Aiden reads "well, that changes everything" three times, producing three different deliveries from the identical text. The only variable changed between requests is the instructions field, with no re-recording or voice cloning involved. This makes delivery style a prompt-controlled parameter rather than a fixed property of the voice.
Qwen3-TTS is Alibaba's (Qwen) text-to-speech model. The workflow shown is simple: the caller sends the same text and same voice selection, and varies only the natural-language instruction attached to the request. The model interprets the instruction to adjust tone, emphasis or pacing while preserving the speaker's timbre, so a single voice can sound sarcastic, earnest or deadpan on demand.
For developers building voice agents, narration pipelines or game dialogue systems, this removes the need to maintain multiple voice presets for different emotional registers. Compared with traditional TTS pipelines that require SSML tagging or separate expressive voice packs, instruction-based delivery control keeps a consistent brand voice while letting the script context dictate the performance.
you can change how a qwen3 tts voice says a line without changing the voice
here aiden reads "well, that changes everything" three times. the words stay the same and we only change the instructions field in the request
listen to all three ↓ https://t.co/UD1VV10DDR
This report is categorized under AITOP's "How-to" desk, focusing on China's AI industry and domestic foundation model ecosystem. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Qwen3-TTS, Together, TTS, prompt engineering.
Source attribution: curated from Together AI (@togethercompute).