MeloTTS English
myshell-ai/MeloTTS-English
published Feb 2024 · updated Dec 2024
MeloTTS English is a text-to-speech model that generates high-quality English speech with multiple accents (American, British, Indian, Australian, and Default).
specs
| Task | Text-to-Speech (TTS) |
| Architecture | VITS-based |
| License | MIT |
| Accents | American, British, Indian, Australian, Default |
about this model
MeloTTS-English is a text-to-speech (TTS) model that generates high-quality, multi-accent English speech. Developed by MIT and MyShell.ai, it supports American, British, Indian, Australian, and a default English accent, each with a dedicated speaker ID.
The model is optimized for low-latency inference and runs in real time on CPU, making it suitable for server-side or edge deployment without GPU requirements. Audio samples for each accent are available in the model card—see the table below for direct links to waveform examples at normal speed.
| Accent | Audio Example |
|---|---|
| American | Listen |
| British | Listen |
| Indian | Listen |
| Australian | Listen |
| Default | Listen |
Key strengths include adjustable speaking speed without quality degradation and a lightweight architecture that achieves CPU real-time inference. The model is built on the VITS/VITS2 family and is released under the MIT license for both commercial and non-commercial use. On the Hugging Face Hub, the model has received over 400 monthly downloads and is used in 20 Spaces as of early 2025.
Developed by Wenliang Zhao and Xumin Yu (Tsinghua University) and Zengyi Qin (MIT, MyShell.ai).
best for
- ·Generating English voiceovers with multiple accents for videos or audiobooks
- ·Real-time speech synthesis on CPU for chatbots and virtual assistants
- ·Adding low-latency TTS to applications without requiring GPU
FAQ
It supports American (EN-US), British (EN-BR), Indian (EN_INDIA), Australian (EN-AU), and a Default accent.
It is released under the MIT License, free for both commercial and non-commercial use.
Yes, the model is fast enough for CPU real-time inference.
Use the gigarouter OpenAI-compatible endpoint with your API key; send a request with the text and desired accent.
This specific model is for English only. The MeloTTS library includes separate models for Spanish, French, Chinese, Japanese, and Korean.
We're benchmarking and onboarding MeloTTS English as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.