Sign InOpen Brain
VercelEngineering PostOfficial Source

Fish Audio models now available on Vercel AI Gateway for free

Vercel AI Gateway added four Fish Audio models for speech generation and transcription, with AI SDK 7 support and a free window whose model naming determines later billing.

Vercel · Aug 19, 2026
Open Source Open MarkdownOpen JSON
Source Summary

AI Gateway added **four Fish Audio models** for text-to-speech and transcription. Regular prices are **$15 per million characters** and **$0.36 per audio hour**, but both are free through September 18.

Practical Implication

Builders can call speech and transcription through **AI SDK 7**. Use standard model names if automatic billing after the offer is acceptable, or append -free to make requests stop when it ends.

Agent-Ready Context
AI Gateway added **four Fish Audio models** for text-to-speech and transcription. Regular prices are **$15 per million characters** and **$0.36 per audio hour**, but both are free through September 18.

Builders can call speech and transcription through **AI SDK 7**. Use standard model names if automatic billing after the offer is acceptable, or append -free to make requests stop when it ends.

The free period is temporary. The material gives capability descriptions but no comparative quality, latency, or reliability measurements for choosing among the models.
Connected Context · Feed7 Judgment

This extends the shared AI SDK and Gateway surface from visual generation into both speech output and transcription, enabling audio pipelines without a separate provider integration. The temporary free routes are useful for bounded evaluation, and the -free suffix uniquely prevents accidental paid continuation; absent comparative measurements, the release establishes access and billing behavior rather than a preferred audio model.

AI Gateway now supports streaming transcriptionStreaming transcription supplies the lower-latency input path that can operationalize Fish Audio’s transcription models; model availability and the streaming transport are complementary prerequisites for live voice-input workflows.Grok Voice Think Fast 2.0 now available on AI GatewayGrok provides an integrated realtime speech-to-speech agent route, whereas Fish Audio exposes separate synthesis and transcription components; choosing between them changes orchestration, latency testing, and credential requirements.Gemini 3.7 Flash now available on AI Gateway for 50% offBoth use temporary Gateway pricing to lower evaluation cost, but Fish Audio’s -free model names can fail closed when the offer ends, avoiding the automatic transition to paid billing that promotion-based routes otherwise require teams to reassess.
Context Map
infraaudio#gateways#agent-sdks#generative-media
Uncertainty
The free period is temporary. The material gives capability descriptions but no comparative quality, latency, or reliability measurements for choosing among the models.