Pictide: an open-source ComfyUI companion that learns your taste as you browse
LykrInSleep · reddit · 2026-10-20
A developer open-sourced Pictide for people who want to enjoy AI image generation without prompt-tweaking. It auto-composes tag-based prompts, sends them to ComfyUI, and lets you swipe through results, liking/disliking as you go. A local recommender learns from that feedback to steer future prompt combinations. Key points:
- Tag combinations account for tag relationships; you can anchor an element (e.g. "forest") and explore around it
- Prompt generation, preference learning and recommendations all run locally, as do image and feedback storage
- Works best with Danbooru-style tag models
Available on GitHub; the author is soliciting feedback on setup and long-term browsing appeal.
More from Multimodal
- Redditor remakes SD clown dance video at 25 steps with motion boost and fixed prompts — roychodraws · 2026-10-21
- Fibo-2, the #1 open image editing model, releases weights on Hugging Face — linoy_tsaban · 2026-10-21
- Ex-Brain Games host made a 25-min AI film in 5 days using 'vibe directing' in iMessage — lmoroney · 2026-10-21
- Motion Launches a General Video Agent That Edits Footage and Mimics Any Reference Style — mhdfaran · 2026-10-21
- MOSS-Image's 5B model stuns with legible-text movie posters that run locally — IcyPickle4540 · 2026-10-21
- What keeps AI videos watchable: hooks, consistency and a payoff — aftahi_ai · 2026-10-21