AI4Bharat releases IndicF5 for 11 Indian languages
AI4Bharat has released IndicF5, a polyglot Text-to-Speech (TTS) model trained on 1417 hours of high-quality speech data. This model supports 11 Indian languages, generating near-human quality speech, and is deployable via Hugging Face.
Key Takeaways
- IndicF5 is trained on 1,417 hours of high-quality speech from Rasa, IndicTTS, LIMMITS, and IndicVoices-R.
- The model supports 11 Indian languages: Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, and Telugu.
- IndicF5 is packaged as a 0.4B-parameter model with F32 tensor type on Hugging Face.
- Example usage requires three inputs: target text, a reference prompt audio, and the transcript of that reference audio.
Why It Matters
IndicF5 gives developers a ready-to-use text-to-speech model for 11 Indian languages, with Hugging Face distribution and example code lowering the barrier to deployment. That matters for localized voice applications that need speaker-conditioned synthesis rather than a single generic voice. The release also adds another open model to the AI4Bharat and Hugging Face ecosystem, with 19,473 downloads last month signaling active interest. Next to watch: whether Hugging Face adds inference-provider support, since the model page says none is deployed yet.
Read full article at huggingface.co