Freefield rebuilds Higgsfield on ComfyUI: 117 apps, open weights only, MIT
neuregex · reddit · 2026-10-28
- For Comfy's #OpenSourceHiggsfield challenge, a developer rebuilt Higgsfield's platform as Freefield (MIT) on ComfyUI and the Comfy SDK.
- 117 apps cover 95 of the 129 tools Higgsfield shows: FLUX 3 Klein image gen, Cinema Studio with 33 camera moves, MiniMax H3 video create/extend, Re-shoot, Relight, 24 LoRA effects, Morph, dubbing, face swap, depth maps, video background removal, brush inpainting, plus 72 template apps like an AI Influencer builder.
- Open weights only, no API nodes: LTX-2.5, FLUX 3 Klein, MiniMax H3, Wan 2.1/2.2, Qwen-Image-Edit, Z-Image, SeedVR2, Stable Audio 3, MiniMax Music 3.
- Each app is one folder (app.json + API-format workflow); the web UI, model downloads, VRAM map, and tests are generated from those files. Any ComfyUI workflow becomes a draft app with one command, and apps chain via "Use in...".
- A lab ran every app with fixed seeds and published real timings and peak VRAM: 48 runs, 710 of 733 cases OK. Runs on Modal (L40S/H100) or any ComfyUI, with 24 GB card guidance in the docs.
More from Multimodal
- Open-source pipeline turns Seedance AI video into real PBR textures on your 3D mesh — OdinLovis · 2026-10-28
- AI-generated cyberpunk short 'VHS Sirens' shows off consistent characters and neon city detail — brythehuman · 2026-10-28
- First-time creator makes full animated music video with 40 clips via MiniMax H3 — Capable-Insurance-58 · 2026-10-28
- Midjourney sref combo 138606468+1306896884 yields textured flat illustration — sergeantsref · 2026-10-28
- Odyssey Publishes Prompting Guide and Full Episode 1 Workflow, From Script to Edit — NicoVerderosa · 2026-10-28
- Striking AI-Generated Starry Wave Image Showcases Current Image Model Quality — ciguleva · 2026-10-28