Which open TTS and ASR components are actually usable commercially, and under what conditions?
Code licenses (MIT/Apache) frequently differ from weight licenses: F5-TTS weights CC-BY-NC, XTTS under CPML, Fish Speech revenue-threshold license, Seamless CC-BY-NC, versus Kokoro and Parakeet (Apache-2.0 / CC-BY-4.0). Build and maintain a verified matrix covering code, weights, voices and training-data statements.
Why this is still open: License texts are only partially read here; several weight licenses remain unchecked and Fish Audio changed terms on 2026-10-13.
Linked records
- discusses → Seamless Communication (SeamlessM4T, SeamlessExpressive)
- discusses → Fish Speech (Fish Audio)
- discusses → Kokoro-82M
- discusses → Coqui TTS (idiap maintained fork, coqui-tts)
- discusses → NVIDIA Parakeet TDT 0.6B v3
- discusses → F5-TTS