Thursday, June 26, 2025 | 4:30 PM - 8:15 PM PDT | GitHub, San Francisco

Join us for an exciting 4-hour hackathon where you'll build custom nodes and workflows that leverage NVIDIA RTX technology to accelerate ComfyUI. With only 3-4 hours of coding time, you'll focus on creating practical solutions that address real performance challenges in the ComfyUI ecosystem. Whether you're optimizing TensorRT for video models, reducing VRAM usage for popular models like Flux, or creating innovative workflows with RTX Remix, this is your opportunity to contribute to the open-source community while showcasing the power of RTX acceleration.

The ComfyUI ecosystem runs primarily on consumer GPUs where every optimization matters. Video generation is currently 10-100x slower than image generation, and popular models like Flux require 24GB+ VRAM, limiting accessibility. By participating in this hackathon, you'll help solve real bottlenecks that affect thousands of users daily. This is a unique opportunity to work directly with NVIDIA RTX technologies including TensorRT, RTX Remix, and various AI SDKs while contributing to one of the most active open-source AI communities. Your solutions could enable new creative possibilities and make advanced AI models accessible to more users.

NVIDIA RTX 5090
Best custom node leveraging RTX technology
*In-person attendance required

NVIDIA RTX 5090
Best workflow showcasing RTX capabilities
*In-person attendance required

$500 Cash Prize
Best RTX-optimized project across all tracks

Surprise box of exclusive Comfy merch and cool stuff for the best virtual submissions! While RTX 5090s are reserved for in-person winners, virtual participants can still win awesome prizes.

Build a ComfyUI custom node, workflow, or extension which either:

• Revolves around optimizing models for NVIDIA hardware using NVIDIA libraries (e.g., TensorRT)
• Uses NVIDIA libraries that work well with ComfyUI like RTX Remix
• Demonstrates how a cool workflow can be optimized with TensorRT, showing the performance boost before and after

The best projects are often original ideas! Use the suggestions below only if you need inspiration.

🔧 Developer Track - Custom Nodes

Focus on building custom nodes that leverage RTX acceleration.

• TensorRT INT8/FP8 Quantization for Flux - Enable Flux to run on 12GB cards
• Video Model TensorRT Accelerator - 2-5x speedup for HunyuanVideo/CogVideoX
• TensorRT ControlNet Support - Add missing ControlNet to ComfyUI_TensorRT
• Smart TensorRT Profile Manager - Auto-detect optimal engine parameters

🎨 Creative Track - Workflows

Build workflows that showcase optimized pipelines.

• RTX Remix Game Reskin Pipeline - AI-enhanced textures
• Cinematic Video Factory - Multi-model pipeline with TensorRT optimization
• Real-time Concept Art Studio - Live drawing to finished art
• Batch Character Consistency - Generate hundreds of consistent variations
• AI Storyboard Generator - Script to visual sequence automation

• 4:30 PM - Doors open
• 5:00 PM - Hackathon introduction, stream starts
• 5:15 PM - Hackathon begins
• 7:15 PM - Hackathon ends, presentations begin
• 7:45 PM - Winners announced
• 8:15 PM - Event ends, Afterparty location announced at event

You can participate as a developer with Python skills, a workflow creator who knows ComfyUI, or a technical creative who can optimize workflows using existing TensorRT nodes and tools. All skill levels welcome!

Projects should leverage TensorRT optimizations, RTX Remix capabilities, or other NVIDIA SDKs - not general UI improvements.

In-person attendees get free Runpod cloud GPU credits. Run TensorRT and GPU workloads even if you only have a laptop!

Set up Python 3.11+, install comfy-cli, test TensorRT nodes, and explore workflows. With only 4 hours to code, preparation is key!

• 🎯 Solo or Teams: Work alone or in teams up to 4 people. Mix of GPU owners and laptop users welcome!
• 🤝 Skill Mixing: We encourage diverse teams (developer + artist + newcomer)
• 📍 Attendance: In-person at GitHub SF or virtual via livestream

We're making this hackathon globally accessible with full virtual participation support.

📺 Watch Live

Stream on Twitter/X and Discord. Get the password during the livestream.

💬 Discord Channel

Dedicated hackathon channel for virtual participants to collaborate and get support.

🎁 Win Prizes

Surprise box of Comfy merch for best virtual submissions!

1. Join the livestream on Twitter/X or Discord at 5:00 PM PDT
2. Listen for the secret password announced during the introduction
3. Build your project during the hacking period
4. Submit via GitHub with the password in your commit message before 7:15 PM PDT
5. Watch the demos and find out if you won!

88 Colin P Kelly Jr St
San Francisco, CA 94107, USA

Questions? Email hello@comfy.org

Register on Luma: https://lu.ma/zndawmg9

Registration is subject to approval by the host

Built with 💜 by the ComfyUI community. Powered by NVIDIA RTX.
Afterparty sponsored by Runpod

• RTX Innovation
  Creative use of TensorRT, RTX Remix, or NVIDIA SDKs
• Performance Impact
  Measurable improvements in speed, memory, or quality
• Community Value
  How many users would benefit from this solution
• Technical Execution
  Code/workflow quality and implementation approach
• Demo/Workflow Quality
  Clear presentation of the problem solved