Fish Audio's Drama 3 text-to-speech model now on Vercel AI Gateway

Vercel says Drama 3, a text-to-speech model from Fish Audio, is now available through AI Gateway and is in preview. The model takes natural-language directions for delivery: emotion and pacing go in square brackets, sounds and pauses can be placed in the script, and selected words can be wrapped in tags. Directions can be written in a different language from the spoken text.

Drama 3 also handles multi-speaker dialogue, with a separate voice for each character. Developers use the model name fish-audio/drama-3 with the AI SDK's generateSpeech function, and the changelog includes a short example that writes an MP3 file to disk. A model playground is also available for trying it.

The 80% discount runs for 30 days, through Nov 12. AI Gateway provides a unified API for calling models, with usage and cost tracking, retries, failover, budgets for API keys and routing rules. Vercel says it passes provider pricing through with no markup and charges no platform fee on inference, including Bring Your Own Key requests.

Teams building voice features can add expressive, multi-speaker speech through one gateway call without a separate integration. Because AI Gateway charges no platform fee on inference, costs follow the model's provider pricing. Drama 3 is still in preview, so check its behavior before relying on it in production.