@anemale.eth
building something new tonight 🧬
I've been working with Claude on a full generative art system
the pipeline :
→ base images processed through ComfyUI (depth maps + canny edge)
→ sent as structural anchors to Replicate + custom LoRA
why the base images ?
because text prompts alone can't fully control spatial composition.
the depth map gives the model a skeleton, volumes, depth planes, horizon logic.
the prompt handles everything else : color, texture, style, emotion.
composition is preserved. everything else is free.
gen art is not just randomness, it's a system with intention.
If you’ve already experimented with ComfyUI, let’s discuss it in the comments.
2 replies
4 recasts
20 reactions