Ming TTS
by inclusionAI · Model · Voice & audio
promising
Open MIT-licensed speech model that ranks near the top of open-weight TTS
Open TTS with voice description, cloning and emotion tags; 3 GB and 18 GB (multi-speaker podcast) models, MIT
AccessOpen weightsNeeds~18 GBDownload3 GB small model; 18 GB multi-speaker podcast modelLicenseMIT
Why we say so
Artificial Analysis ranks the smaller Ming 3.0 model among the best open-weight text-to-speech models, slightly ahead of Fish Audio's Drama 3. Ant Group adds voice cloning, voice description and emotion tags. That is one independent benchmark and no hands-on reports yet.
Evidence
- Oct 25Smaller model ranks among best open-weight TTS on Artificial Analysis, only slightly above Fish Audio's Drama 3