Indic TTS

An open-source text-to-speech toolkit generates speech in Indian languages with downloadable models and code for local inference and training.

Alternative to:

ElevenLabs

Indic TTS is a text-to-speech toolkit from AI4Bharat for developers and researchers building speech applications in Indian languages. It provides trained models and training code for generating speech on your own hardware.

Indian language speech: Generate audio in Assamese, Bengali, Bodo, Gujarati, Hindi, Kannada, Malayalam, Manipuri, Marathi, Odia, Rajasthani, Tamil and Telugu.
Downloadable models: Use pretrained model weights and configuration files for speech synthesis without training a model yourself.
Speech synthesis architecture: The language-specific models pair FastPitch with HiFi-GAN and use joint training on male and female speakers.
Training and data preparation: Train and test models with the supplied code, prepare IndicTTS data in LJSpeech format, and assess the dataset's suitability for speech synthesis.

The toolkit is open source under the MIT license. Local inference uses downloaded model files through a Python workflow based on Coqui TTS. The training setup uses PyTorch and CUDA GPUs; this is a code-based toolkit for application development and research.

Fine-Tuning and Training Text-to-Speech Models Text-to-Speech and Audiobooks Linux Multilingual