Talk Beginner MIT

Building Voice Agents using FOSS models

Rejected

Session Description

Whenever we think of Voice cloning or voice agents we get Elevenlabs, but all these are possible locally on your desktop/laptop/macbook. In this talk I will showcase how to build your own Voice clone agent using 3seconds to 10seconds of audio and Opensource LLM Models such as Qwen, Llama with TTS and STT which runs smoothly on our i3 machines too.

Key Takeaways

Takeaways:

Understanding of Text to Speech (TTS)
Workings of Speech to Text (STT)
How Voice cloning works, which models are opensource and why to use them
How LLM places a roles here for JARVIS persona

References

https://dishantg.com/#speaking

Session Categories

Knowledge Commons (Open Hardware, Open Science, Open Data etc.)

Technology architecture

Talk License: MIT

Speakers

Dishant Gandhi AI Consultant | Encoura

AI Consultant, building AI agents for the last 6years for Overseas clients and Public speaker for at least 50x across the globe.

https://www.linkedin.com/in/dishant-gandhi/

Reviews

Thanks for your submission. We are passing on this proposal. Local voice cloning and pipeline setups are pretty well trodden ground at this point, and the takeaways read more like a general guide than a deep technical session. We want talks with deeper architectural insights or unique challenges solved, so we cannot offer a slot this time around.

Reviewer #1 Rejected