DEV Community

Cover image for Real-time text to speech conversation about friends and Ray Bradbury with my computer
w4ffl35
w4ffl35

Posted on

Real-time text to speech conversation about friends and Ray Bradbury with my computer

About this update

This video shows a conversation between myself and the AI. It uses SPD in order to achieve real-time text to speech generation. Previous videos have showed off TTS using Bark and SpeechT5 which sound more natural but take more processing power. Sentences generate fairly quickly with those models, but they can take upwards of 19 seconds to generate in the case of Bark or 2 seconds in the case of SpeechT5. This causes the conversation to lag and sound less natural.
spd is a build in system speech synthesizer so it sounds robotic, but its very fast. This makes the conversation feel more natural even though it sounds less-so.

This video also showcases updates to the drawing tools and I talk about over all performance improvements which I have made to the app.


More information about AI Runner

AI Runner is an application that I have been developing for over a year. It is a multi-modal AI app which uses Huggingface libraries to serve up AI models locally from your own computer (no internet connection required after you've downloaded the models).

It gives you a ton of control over the content you can generate, uses Stable Diffusion for AI Art and more.

The upcoming 3.0.0 version has been in development for around 7 months and aims to evolve the app from a Stable Diffusion art application into a multi-modal AI assistant.

AI Runner is free. The compiled 2.2.0 version can be downloaded here.

The source code is here


Supporting me on Patreon will help to ensure development of AI Runner continues into the future.

Top comments (0)