Higgsfield promised to open-source their entire platform. They didn’t.
Let's help them.
For the next two weeks, Comfy and our partners are putting up $25,000 in prizes for the best open-source builds on the Comfy Developer Platform.
Partners: @NVIDIAAI, @runpod, @MiniMax_AI, @ltx_io, @bfl_ai
→ $10K grand prize
→ RTX 5090: Most Innovative
→ RTX 5090: Most Viral
→ RTX 5090: Open Weights Only
→ 10 runners-up: $250 in Comfy credits
The studios that win with AI won’t have the best model.
They’ll own their pipeline.
We went through 450+ customer calls this year. The top reason teams pick @ComfyUI: control over every step.
Only 22% asked for it.
You don’t ask for what you already rely on.
If you're using AI in production, do you own your pipeline or rent it?
@HeyGen Avatar 6 is now available in ComfyUI via Partner Nodes.
Give it a face and a voice. Get back a full on-camera performance.
→ Lip sync, gestures, expression and posture in one take
→ Direct it in plain language with a motion prompt
→ Photo, reference video or HeyGen avatar as input
→ 1080p or 2K, five aspect ratios
→ Runs locally or on Comfy Cloud
GTA, but set in prehistoric times 🦖
Turning a simple 3D blockout into a prehistoric GTA-style scene with ComfyUI.
Prompt + workflow included in link below 👇
MiniMax H3’s video VAE is ~2x faster in ComfyUI on @nvidia GPUs
→ Encode: up to ~2.2x faster
→ Decode: ~1.4–2.7x faster
→ 1344x768, 129-frame round trip: 24.3s → 12.7s
What changed under the hood:
• A fused encoder kernel removes redundant memory passes between convs (on by default, lossless)
• A custom conv lets the encoder use fp16 accumulation
• The int8 decoder keeps intermediate results out of memory
Quality holds: the added error sits far below the VAE’s own compression loss.
To get it, update to v0.36.0+, launch with --fast fp16_accumulation, and load the int8 H3 VAE.
Full breakdown from @Kijaidesign in our blog ⬇️
@bria_ai_ RMBG 3.0 is now in ComfyUI via Partner Nodes.
→ Cleaner cutouts on hair, fur and motion blur
→ True 8-bit alpha matte, not a hard mask
→ Same image, same result, every time
→ Trained only on licensed dataSelect RMBG3 in the Bria Remove Image Background node.
Seedance 2.5 Flash is now in ComfyUI via Partner Nodes.
→ Outperforms Seedance 2.0 at a lower cost than Seedance 2.5
→ Text, first/last frame, and multi-reference
→ Edit or extend reference videos
→ 4–30s at 720p with native audio
Click the link below to try it today ⬇️
I'm yet to find a limit to what's possible with custom nodes in @ComfyUI..
I'm working on a group of nodes that pull frames from a clip, give it to an LLM that describes what it sees and hands it to another set of nodes that animate + track the type onto the clip. Super fun!