Long videos with full control - Wan 2.2 I2V (SVI 2 Pro). Individual LoRa, multiple reference images and more

This workflow focused on ease of use and powerful features for creating segmented videos.

Features:

• Create unlimited videos beyond Wan 2.2 capabilities, thanks to SVI 2.0 pro. Easily add new sections to continue your video further, all in single continuous workflow

• Automated save/loads. Continue your video from where you stopped, even after comfy restart. No quality degradation on video compression

• Change base models for each section as well as LoRa. Find which ones work best for your idea

• GGUF support for low VRAM systems. Use your GGUF clip and diffusion models like usual models at unified fields

• Separable upscale/interpolation at the end. You can stitch a full video purely from saved data, preserving RAM and VRAM, or just generate from the beginning to upscaled video in one go

• RAM optimizations. If your system can generate 1 segment, you should be able to do them all!

• Chose different reference images for each video segment. It can introduce model to hidden assets/characters, that aren't visible on the start image. No LoRa required!

• Precisely control every video segment: prompt, length, seed. Check for perfect stitch with previous generation only, no need to search through entire video

You can find guide for the workflow inside Options subgraph (v3.0), it's your starting point.

Feedback is appreciated! Share your thoughts

Version 3.0

See changes in version history

For this version you'll need my new custom nodes. If comfy manager can't recognize them automatically, you can find them manually by name:
ComfyUI-HeresyNodes

Or download them using git repository link
https://github.com/heresyCoder/ComfyUI-HeresyNodes

Remove old nodes ComfyUI-WanVideo-LatentIO used in V2 if they cause conflicts, they are added in the new node pack with some changes

April and Kit example was made using multiple input images. You can find them in the image gallery of v3.0

Note that workflow in this example was upscaled separately from generation, that's why all sections are bypassed.
If you want to use it, uncheck full_video_only parameter on options subgraph, unbypass sections and turn off upscale/interpolation/save on full video subgraph.

Or just download as JSON file

Version 2.0
See changes in version history

Based on kijai's SVI 2.0 workflow:

https://github.com/user-attachments/files/24359648/wan22_SVI_Pro_native_example_KJ.json

New custom nodes were made for saving latents of a video (Save Wan Video Latent) and loading from "output" folder (Load Wan Video Latent).
If you can't install them with comfy manager, you can get them manually:

https://github.com/heresyCoder/ComfyUI-WanVideo-LatentIO

Frieren with a mimic example was made using multiple input images. You can find them in the image gallery of v2.0