Fish Audio Drama 3 text-to-speech is 80% off on Vercel AI Gateway at $10 per million characters through November 12

Fish Audio released Drama 3, its new text-to-speech model, on October 13, and Vercel AI Gateway added it the same day. For the first 30 days the Drama 3 API is 80% off, which makes this a good window to try it for voiceovers, character dialogue, and audio storytelling.

What you get

Through November 12, Drama 3 on Vercel AI Gateway costs $10 per million characters, down from the $50 list price. AI Gateway passes provider pricing through with no markup and no platform fee on inference, so the discounted rate is what you pay. Any Vercel user with AI Gateway access can use it, with no coupon or promo code needed.

Drama 3 is currently in preview. The same discounted rate applies on Fish Audio’s own API, and new Fish Audio creator plans are 50% off over the same period.

Specs and public benchmarks

You direct Drama 3 in plain language. Put emotion and pacing in square brackets, such as "[barely holding it together, forcing a smile]", and breaths, pauses, and laughs land where you write them. To change how only a few words are read, wrap them in a paired tag. Directions can be in a different language from the spoken line, the model speaks 83 languages, and it supports multi-speaker dialogue with a separate voice for each character.

On Artificial Analysis’ Pronunciation Robustness benchmark, Drama 3 read 93.2% of hard words correctly, the top score on the leaderboard, compared with 88.1% for S2.1 Pro. Accuracy on brand names, places, and technical terms rose from 87.3% to 95.9%, words whose reading depends on context went from 86.1% to 95.3%, and exact sequences such as codes, file paths, and emails improved from 79.4% to 86.3%.

The weights are on Hugging Face, free for commercial use if you and your affiliates make less than US$500,000 a year.

How to use

1. Sign in to Vercel and create an AI Gateway API key.
2. Call it with the AI SDK generateSpeech function and set the model to fish-audio/drama-3. For the voice parameter, open a voice in the Fish Audio voice library and copy its Model Id from the menu.
3. To hear it first, use the playground on the Drama 3 model page.

import { experimental_generateSpeech as generateSpeech } from 'ai';

const result = await generateSpeech({
  model: 'fish-audio/drama-3',
  text: '[warm and reassuring] Take your time. <whisper>I am right here.</whisper>',
  voice: 'f3a2b90078d54a65af4b95f63bc7798e',
  outputFormat: 'mp3',
});

See the Vercel changelog post, the AI Gateway docs, and Fish Audio’s Drama 3 launch post.