Preview for FramePack on ComfyUI

Run this workflow on InstaSD

Get started in minutes! Run this ComfyUI workflow online - no setup required.

Description

Unlock the power of video generation with the Frame Pack Workflow on ComfyUI, powered by InstaSD. Based on the innovative Frame Pack method by Llyasviel and wrapped by Kijai, this workflow allows you to create seamless videos by generating multiple frames in a single pass. Ideal for animators, video creators, and AI enthusiasts, this setup provides smooth video outputs by leveraging the power of next-frame prediction and optimized processing.

🎯 Features

  • Efficient Video Generation – Generate smooth, high-quality video frames at 30fps.
  • Next-Frame Prediction – Uses advanced techniques for predicting and diffusing the next frame in a sequence.
  • Anti-Drifting Sampling – Reduce quality degradation in long video generation sequences, ensuring stable output.
  • Cloud-Powered Performance – Run the workflow on InstaSD’s high-VRAM GPUs, optimized for video processing.

đź’ˇ Use Cases

  • Animation & Video Production – Create fluid and professional-quality video sequences.
  • AI Research & Experimentation – Explore advanced video generation models and test next-frame prediction methods.
  • Creative Video Projects – Generate video content for digital art, concept development, or media production.

⚙️ How It Works

  1. Launch the Workflow – Open the preconfigured ComfyUI instance with Frame Pack.
  2. Customize Your Frames – Input your prompt and adjust frame settings for the video output.
  3. Generate & Preview – InstaSD’s cloud infrastructure handles the heavy processing for seamless video creation.
  4. Download Your Video – Export your generated video frames or complete video for use in projects or further editing.

Models

FileDestinationSource
FramePackI2V_HY_fp8_e4m3fn.safetensors/ComfyUI/models/diffusion_modelsDownload
sigclip_vision_patch14_384.safetensors/ComfyUI/models/clip_visionDownload
llava_llama3_fp16.safetensors/ComfyUI/models/text_encodersDownload
clip_l.safetensors/ComfyUI/models/clipDownload
hunyuan_video_vae_bf16.safetensors/ComfyUI/models/vaeDownload
250413_165330_476_4350.png/ComfyUI/inputDownload

Nodes

VAEDecodeTiledVAEEncodeCLIPVisionEncodeFramePackTorchCompileSettingsGetImageSizeAndCountConditioningZeroOutDownloadAndLoadFramePackModelLoadFramePackModelCLIPVisionLoaderMarkdownNoteDualCLIPLoaderVAELoaderCLIPTextEncodeFramePackSamplerImageResize+VHS_VideoCombineFramePackFindNearestBucketLoadImage