Fish Audio Drama 3 Launches Directed Voice Control
Fish Audio’s new text-to-speech model turns scripts into directed performances using natural-language cues for emotion, pacing, whispers, laughter, and pauses. Drama 3 supports 83 languages across web, mobile, and API, with open weights for self-hosting.
Drama 3’s real breakthrough is making voice direction scriptable rather than preset-driven, giving developers finer control over authored audio. Its open weights are compelling, but licensing limits and demanding hardware make it less turnkey than hosted alternatives.
- –Mid-sentence direction changes enable more convincing audiobooks, game dialogue, podcasts, and character scenes.
- –The same API endpoint as S2.1 Pro should make migration relatively straightforward for existing Fish Audio integrations.
- –Open weights expand customization and self-hosting options, but the community license is not fully open source and restricts larger commercial deployments.
- –Its 35B mixture-of-experts architecture has roughly 3B active parameters, but self-hosting still requires substantial GPU memory.
- –Variable takes and unfinished editing tools mean production teams may still need manual review and regeneration.
DISCOVERED
1h ago
2026-10-16
PUBLISHED
9h ago
2026-10-16
RELEVANCE
AUTHOR
[REDACTED]