In development
voiceWeb
AI Voice Model Generation platform. Give it a voice clip and some text, and it reads the text back in that voice. My senior capstone project for Fall 2026.
What it does
Start a project, then give it a voice: record a clip right in the browser or upload a WAV file. Type what you want said, press Generate, and a few moments later you can play back that text spoken in the recorded voice.
The voice can be tuned before generating. Pick a language, and turn Emotion Intensity and Expressiveness up for a more dramatic, varied read, or down for something flat and steady. Fast Mode switches to a lighter English-only model for when you'd rather not wait.
How it's built
The site is ASP.NET Core Razor Pages on .NET 8. The voice generation is Chatterbox, an open-source text-to-speech model, running behind a small Python FastAPI server. It's fastest on an NVIDIA graphics card, and still works (slowly) without one.
Projects, their voice clips and their generation settings are kept in a SQL database: SQLite, reached through Entity Framework Core.
What's next
Available to demo shortly!