Playground
Things I built for fun. Some still run in your browser. The AI tools ran on a GPU at my place in California, and they've been offline since I moved to New York, so each one has a write-up of how it worked instead.
works right now
- Pitch checker live Real-time pitch detection for voice and guitar. Runs entirely in your browser.
- MIDI studio live A small music workstation in the browser. Piano roll, synths, drums, effects, WAV export.
- Claude Corner live Once a month Claude writes something for this site. I read it before it goes up.
ran on my GPU
- austn.net's GPU tools case study How all of these shared one GPU, what broke, and what I changed.
- Image generation Text to image with Stable Diffusion 1.5, run through ComfyUI. offline · Stable Diffusion 1.5 (fp16) on ComfyUI
- Music generation Full songs with vocals from lyrics and style tags, using ACE-Step. offline · ACE-Step on ComfyUI
- Text to speech Speech from text with Chatterbox, with adjustable delivery and share links. offline · Chatterbox, on its own Flask server
- Stem separation Split a song into vocals, drums, bass and everything else, with Demucs. offline · Demucs (htdemucs, htdemucs_ft) on ComfyUI
- Background removal Upload an image, get a transparent PNG. Seven models for different kinds of pictures. offline · rembg on ComfyUI (u2net family, ISNet)
- Image to SVG Trace a raster image into a scalable vector with VTracer. offline · VTracer on ComfyUI
- Image to 3D Turn a photo of one object into a 3D model you can spin around in the browser, with Hunyuan3D. offline · Hunyuan3D 2.1 on ComfyUI
- Chat A chat box backed by a local model in LM Studio. Slow, but free and private. offline · Qwen 2.5 Coder 14B in LM Studio