Show HN: Wall-mounted diffusion mirror that turns reflections into paintings

Base64 encoding the JPEG bytes will increase payload size up to ~30% and burns CPU cycles on both client and server. This is unnecessary, as Websocket protocol can send binary payloads (doesn’t need to be text).

Consider removing lossy jpg compression as well, ie just send the raw RGB bytes over the network. Then on the server side you can simply call Image.frombuffer(…).

StreamDiffusion can achieve high frame rates because of extensive batching in the pipeline. You’re not benefiting from that here as the client is only sending one frame at a time and then waiting for a response. See this example for an idea of how to queue input frames and consume them in batches https://github.com/cumulo-autumn/StreamDiffusion/blob/main/e... .

Alternatively you could take a look at the SDXL Turbo and Lightning models. They are very fast at img2img but have limited resolution of 512² or 1024² pixels respectively. Which appears a bit lower than what you’re aiming for here, but they can be run locally in real time on a high end consumer grade GPU. For reference I have some code demonstrating this here https://github.com/GradientSurfer/Draw2Img/tree/main

> The main issue with the current setup is the low frame rate.

I'd call it a feature rather a limitation, it's not bad that I have second to process the image for a bit.

I would even increase the update frequency to 5 - 15 minutes and let it capture and generate a new image whenever it detects something changing / moving.