Fish Audio

Fish Audio, at fish.audio, is an AI voice platform for text to speech, voice cloning and speech to text. You type a script, pick a voice, and steer the delivery with tags such as [whispering], [excited], [laughing] or [long pause]. The site says its voice library holds more than 2,000,000 voices and that it supports 30+ languages.

Voice cloning works from a short recording. The home page says 15 seconds and its FAQ says as little as 10 seconds. Cloned and library voices can be used for video voiceovers, audiobooks, character voices and chatbots. The platform also lists speech to text, a voice changer, audio separation, audio translation and sound effects, along with image, video and avatar generators that lip-sync to a Fish Audio voice.

Fish Agent is the company's platform for real-time voice agents. You set a prompt, a voice, a knowledge base and tools in a no-code builder, test the agent in the browser, and publish it to a website widget, your app, the API or a phone number. It can run on Live 3, a full-duplex model that keeps listening while it speaks, and that speaks English for now. Fish Audio's docs describe Agents as a public beta.

Developers get a REST API and SDKs with pay-as-you-go credit that is billed separately from subscriptions. Fish Audio has also open-sourced speech models, including Fish Speech and S2. The company behind it is Hanabi AI Inc., and its co-founders are Rissa Cao, Shijia Liao and Jiahua Liu.

Pricing: Freemium. The Free tier has 8,000 credits a month (up to 7 minutes of generation) for personal, non-commercial use. Plus is $15 a month, or $11 a month billed annually, with 250,000 credits. Pro is $100 a month, or $75 a month billed annually, with 2,000,000 credits and 3 team seats. The API is pay-as-you-go and billed separately. Voice agents are listed at $0.035 a minute through December 31, 2026 (regularly $0.07), plus $0.01 a minute for phone calls and $1.20 a month for a phone number, with a $5 starting credit. Checked November 12, 2026 at https://fish.audio/plan/.