Fish Audio Models Available for Free on Vercel AI Gateway for 30 Days

Fish Audio's audio models have been officially launched on Vercel AI Gateway. To celebrate the launch, all models are free until September 18; after that, text-to-speech (TTS) will cost $15.00 per million characters and speech-to-text (STT) will cost $0.36 per hour of audio.

The four available models are:

• fish-audio/s2.1-pro: Optimized for low-latency streaming, allowing voice cloning via reference recordings.
• fish-audio/transcribe-1: A transcription model that returns text with audio duration and word-level timestamps.
• fish-audio/s2-pro: Supports approximately 80 languages and allows changing the speaking style of specific words via inline tags in the text.
• fish-audio/s1: A TTS model that reads text containing emotion, tone, and sound effect markers.

During the free period, using the model names as-is will trigger automatic billing after the period ends; if you do not want to be billed, you must append the -free suffix to the model name (e.g., fish-audio/s2.1-pro-free). These models are supported in the AI SDK 7 release and can be easily utilized via the generateSpeech and transcribe functions.