Quick correction for anyone reading: SDXL and ZImage weights are 8GB+ in fp16, and WH uses Soulkyn clusters for image, text, and sound generation — none of it runs locally. 300MB of RAM cannot run either model. Full stop. What you’re seeing is normal Chromium webview overhead, standard for desktop apps built this way and unrelated to image generation.
You came into our Discord opening with an accusation that we were secretly using your hardware, with no hello, no question, just the conclusion. I told you that was wrong and called it as I saw it (idiotic). I stand by that. If you don’t want to install the app, the web version exists and requires nothing on your machine. Either option is fine, but the claim doesn’t hold up.
And frankly, this kind of baseless accusation is actively harmful. Plenty of projects in the AI space really are doing shady things with people’s hardware, and it’s easily verifiable when they are. Throwing the same accusation at projects that aren’t doing it makes the real cases harder to call out. Good job.
And since i’m correcting stuff… “I called the accusation idiotic” ≠ “I called you stupid.”, but apparently even this is too complicated. (See there it was you, notice the difference?)