Fish Audio Alternatives

Direct alternatives to Fish Audio

Fish Audio pairs a hosted text-to-speech and voice cloning API with Fish Speech, its model on GitHub (32.8k stars as of September 2026). The code and weights ship under the Fish Audio Research License rather than a standard open-source licence. People comparing alternatives usually want weights under a permissive licence, a different hosted API, or lower latency for agents; the list is grouped that way.

Permissively licensed weights

• VoxCPM: OpenBMB's TTS model under Apache 2.0, covering 30 languages with voice cloning from reference audio and voice design from a text description. The 2B-parameter model needs roughly 8GB of VRAM.
• KugelAudio: publishes kugelaudio-open under MIT, a TTS model for European languages with voice cloning. The same Berlin team runs an EU-hosted API billed per generated audio minute.

Hosted voice cloning APIs

• ElevenLabs: TTS, voice cloning, dubbing and speech-to-text from one API. There is a free option; the Starter plan is $6/month as of September 2026.
• Respeecher: Respeecher Space is a real-time TTS API priced at $2 per hour of audio, pay-as-you-go. Respeecher also does speech-to-speech conversion, re-voicing a recorded performance with voices licensed from real people.

Low-latency speech for agents

• Cartesia: Sonic TTS with 44 languages on Sonic 3.6 (September 2026), deployable in the cloud, on-premise or on-device.
• Rime AI: conversational TTS with cloud, VPC and on-premise deployment. A fit when the voice has to run inside your own environment.