Skip to content
UsecaseBoard
← All use cases
blogTTS

Fish Audio Drama 3: open-weight 35B MoE TTS you direct in brackets

At a glance

Fish Audio (Oct 13) ships Drama 3 — 35B MoE (~3B active) emotive TTS with square-bracket stage directions, 83 languages, #1 on Artificial Analysis Pronunciation Robustness (93.2%). Open weights on Hugging Face under Fish Audio Community License (free commercial under $500k revenue); API via model: drama-3. Built for audiobooks, audio drama, game dialogue, and short video; interactive editing and audio inpainting in invite-only preview. This workflow is listed under TTS, open-weight, voice, multilingual, creative-AI.

Agents / tools
Drama 3, Fish Audio, TTS, voice agents

Building a tool for these workflows? Sponsor / inquire.

Save ideas in this browser. Upvotes count across the board.

Prompt excerpt

POST api.fish.audio/v1/tts with header model: drama-3 and text like "[through gritted teeth] Just one thing. [whispering] Do not open that door."
Explore all →
docscoding-agent

Reflection Mirror CLI: terminal harness optimized for Beam

Mirror is Reflection's early-beta coding agent for the terminal—defaults to Beam-1.0-501B-A23B, manual/auto tool approval, resumable conversations, parallel /worktree agents, AGENTS.md+skills+MCP/extensions, with provider keys kept out of the agent's shell environment.

Mirror CLIBeamReflection+2
Source ↗