dwani.ai is a self-hosted GenAI platform designed to provide voice mode interaction for Kannada and other Indian languages.
We provide a python-sdk to get started with the dwani.ai project. (pip install dwani)
Offline modules for Automatic Speech Recognition(ASR), Text to Speech(TTS), Translation, Document Query and LLM can be used independently with the corresponding repository on the dwani,ai : GitHub Org : https://github.com/dwani-ai
We have integrated the Open-weight models for ASR/TTS/Translate provided by AI4Bharat from IIT Madras, OlmOCR for document queries and Google Gemma3 for LLM/VLM
Start building applications for Indian languages as core feature.
Use the Speech-to-Speech module to run a complete Voice Assistant or use the individual modules
in your project based on feature requirement.
Ex. Using Vision Language Model + Translation + TTS : To assist Visually challenged users with voice description of the immediate environment in local language like Kannada
Would be better as a talk. Especially Q&A about the process of creating this and how more Indic language AIs could be created
A bit of focus on data, its collection and openness of it will make it better suited for this track.
A lightning talk to introduce the project than a workshop would be better.
May be rework on the proposal?