Fish Audio Drama 3 vs ElevenLabs v4, Gemini 3.8 Flash TTS and Grok Voice

Drama 3 is Fish Audio's text-to-speech model for directed voice performances. You describe in plain language how a line should be delivered, using bracketed cues like [sigh], and paired tags can limit a delivery to chosen words. Directions can be written in any language, and it speaks 83. It is a 35B mixture-of-experts model with about 3B active parameters.

Pricing (per 1K characters):

• Drama 3: $0.002 during the launch promotion (80% off, until November 12, 2026). Fish Audio's list price for its own API is $0.05.
• During the promotion it is 7.5× cheaper than Fish's own S2.1 Pro ($0.015) and 40× cheaper than Eleven v4 ($0.08). Check the price again after the promotion ends.
• Eleven v4 Turbo $0.04 · Grok Voice TTS 1.0 $0.015
• Gemini 3.8 Flash TTS is cheaper still at $0.0005, but it is not built for performance direction.

Quality signals: Fish Audio reports 93.2% on Artificial Analysis' Pronunciation Robustness benchmark (English), against 88.1% for S2.1 Pro. The biggest gains are on brands, places and technical terms (95.9% vs 87.3%).

Caveats: Drama 3 is labelled a preview. The weights are open under the Fish Audio Community License, which is free for research and for commercial use below $500K annual revenue. Offering it as a TTS or voice-cloning service needs a license from Fish Audio.

Drama 3 (Fish Audio) · Launch post