Fish Audio's open-weight Drama 3 TTS lets you direct voices word-by-word like an actor
EXM7777 · x · 2026-10-14
Fish Audio launched Drama 3, which it calls the most emotive and controllable TTS model available — and it's open-weight. The company raised a $52M seed round and claims 15M+ users.
Key capabilities:
- Natural-language direction of delivery ("say it like you've given up on life")
- Word-level editing: fix a single word while keeping the rest of the take intact
- Mid-sentence voice switching and multi-character voicing from one script
- Voice cloning that stays in character across languages, demoed by having Uncle Roger roast his own fried rice
It's a direct shot at ElevenLabs' voice business, with open weights as the differentiator.
Related event: Fish Audio Open-Sources Drama 3 TTS Model(2 posts)→
More from Multimodal
- Gradium TTS lands on Baseten with 50ms TTFA for full voice agent pipelines — mattturck · 2026-10-14
- This anime clip cost $1.7 and took 3 minutes; Higgsfield offers unlimited Wan 3.0 for 7 days — piotrbinkowski · 2026-10-14
- AI-Generated Arthur Morgan (RDR2) Video Gains Traction on Reddit — Euratza2052 · 2026-10-14
- AI Sentai Monster Fight: Dragon Ronin Dinozon VS Bus Yarō — RobotMonsterArtist · 2026-10-14
- AI-generated F1 car zipper pull design looks like a real zinc alloy product render — Delahuntagram · 2026-10-14
- Pixel-Curator: Local CLI That Turns 1,000s of Raw Images Into LoRA-Ready Datasets — Impossible-Car-2809 · 2026-10-14