Qwen3 TTS: change how a voice delivers a line via the instructions field, no voice swap needed

togethercompute · x · 2026-11-23

Together AI demonstrates a practical Qwen3 TTS technique: keep the same voice and change only the instructions field in the request, and the same line "well, that changes everything" is delivered with three different readings while the words stay identical. This offers a simple way to precisely control tone and emotion for dubbing and podcast workflows.

Related event: Together AI Launches Qwen3 TTS with Instruction-Controlled Tone(2 posts)→

Original post →

More from Multimodal

Multimodal channel →