prism

-progress is slow just came back from the hospital

Was able to add a way for you to demo it in your browser

-added a fastapi endpoint that runs the denoiser on the uploaded audio
-wired the apps webpage to the api so people can try it without having to install the app

the app is only availbale to use on windows btw (linux and mac support coming soon)

added this

to the site so users can understand how it works

i added ultra‑light GTCRN model support

i updated the website to show v0.1.0 available for download

created a website for prism where people can download this project and read more about it(see the roadmap, read release notes)

added MIT license
created an executable(.exe) for prism
dropped v0.1.0 ,it’s on the GitHub releases pages

• Added DeepFilterNet (ONNX) as an optional denoiser, swappable
  via config alongside RNNoise
• Added adjustable noise removal slider (0-100%) with separate
  on/off toggle
• Added noise meter: shows how much noise is being removed +
  estimated room noise floor

phase 2 done I guess 🎉

will improve the ui
plan to use tauri later on but for now tkinter will do

Built Phase 2 (features i added)

• RNNoise AI noise removal
• Fixed the WO Mic crash
• Fixed your voice getting clipped at the start of words

built a minimal Tkinter UI for Prism: a small window with an on/off toggle for filtering, a dropdown to switch between microphones, a live level meter with gate status, and a status line. To support it, the audio code was refactored from a blocking loop into a controllable AudioEngine that can start, stop, and switch mics at runtime — a stopgap until the Tauri UI, but the engine part is permanent.

currently i’m in my hostel and there is a lot of noise in the background

the images show a before and after the filtering of noise

Added safe mic auto-detection. Prism now validates the input device instead of blindly using the Windows default — it rejects VB-Cable and Sound Mapper pseudo-devices (which cause a feedback loop where Prism listens to its own output), falls back to the first real microphone, and exits with a clear help message if no mic exists.

no ui yet 🤧

One afternoon I was giving a presentation in my rooom, and the room came with me , a ceiling fan , my roomate watching tiktoks in the background, the low hum a room always has. All of it rode my mic straight into the call. I couldn’t mute (I was the one talking), and I couldn’t fix the room.

So the idea for Prism was simple and a little stubborn: your voice should be the only thing that leaves your microphone

progress so far:
-work on the project structure,
-Capture physical mic input
-Route processed audio to virtual cable output (VB-Audio)
-High-pass filter (the processing layer for now)