Open-weight speech models you can run yourself.
Fish Audio develops open text-to-speech and voice-cloning models with multilingual support\, released openly and usable through a hosted API or self-hosted entirely.
Most quality voice synthesis is closed and metered per character; open weights change the economics completely if you generate at volume. Quality trails the leading commercial models\, and open cloning capability makes consent handling your responsibility.
Quick Information
Platform
Web
Pricing
Free open models + hosted API
API
Available
Category
Voice
Pros and Cons
Pros
- Open weights you can self-host
- No per-character cost when self-hosted
- Multilingual support
- Voice cloning included
- Hosted API also available
Cons
- Quality trails leading commercial models
- Self-hosting needs GPU capability
- Open cloning raises consent obligations
- Documentation is thinner