AI BriefingNeuTTS-2E, emotion-controllable on-device TTS open-sourced
Key point
Neuphonic has open-sourced NeuTTS-2E, a 125M-parameter on-device TTS model capable of controlling 7 emotions.
Details
Neuphonic has open-sourced NeuTTS-2E, an on-device TTS model with emotion control capability, as an alpha version.
Key features:
- Lightweight design with 125M active parameters
- Ability to directly specify 7 emotions including angry, fearful, happy (explicit instruction method rather than inferring meaning from text)
- Support for 4 built-in voices, maintaining speaker identity when changing emotions
- Fully local execution — no text or audio sent externally
The company stated that during model training, it addressed issues such as lack of emotional voice data, label reliability problems, and disentanglement between speech emotion and text meaning.
The GitHub, Hugging Face model collection, and interactive demo are all publicly available, and the model can be used under the NeuTTS Open License.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Jevper Released: A Jev Classification Wrapper for OpenAI-Compatible Models with Automatic Logprobs Detection
Rootless Linux Sandbox 'Drop' Released for Isolating AI Agents and Third-Party Programs