Overview
This workflow uses Wan2.2 I2V together with SVI 2.0 PRO LoRA to generate long-form videos of 81 frames or more.
Instead of generating 81+ frames in a single pass, the video is created in multiple segments and then merged.
Compared to traditional extend/long workflows, this approach preserves motion information from the previous segment, resulting in much smoother transitions without noticeable seams.
Core Features
Lightx2v-only Mode
Offers high prompt adherence and fast generation speed, but may introduce the characteristic slow-motion effect typical of Lightx2v.
3-pass Mode
Slower generation speed, but capable of producing more dynamic and realistic motion in real time.
Optional Features
Resume Mode
Allows you to continue generation not only from the first frame, but also from a previously generated video.
If something goes wrong partway through the process, you can save the generated segment and resume from that point without starting over.
Recommended Settings (Updated: 2026-01-14)
Use 3-pass mode with the following settings:
Normal step count: 1
Lightx2v step count: 2
CFG: 2.5
Lightx2v LoRA:
wan2.2_i2v_lightx2v_4steps_lora_v1_xxxx_noise (comfy native)
High: 1.0 / Low: 1.0
I tested many different configurations, and this setup produced the most dynamic motion while maintaining natural motion speed and avoiding artifacts.
Important Tip for Reducing Slow Motion Caused by Lightx2v LoRA
To reduce the slow-motion effect introduced by Lightx2v LoRA, it is crucial to describe motion in detail within your prompt.
Simply writing something like “a woman is running” is not sufficient.
Describing the motion step-by-step in a time-sequenced manner greatly increases the chance of avoiding slow motion and achieving faster, more dynamic movement.
❌ Bad Example
A Japanese woman in a black hoodie is running at full speed in front of her.
The camera follows her movement from the front.✅ Good Example
0-1s: A Japanese woman in a black hoodie starts running forward at full speed
1-2s: The camera follows her movement from the front
2-3s: The woman accelerates her running motion
3-4s: The camera maintains a steady follow shot from the front
4-5s: The woman continues running at high speed with the camera still following from the frontBy explicitly defining temporal progression and motion changes, Lightx2v is more likely to generate natural-speed movement instead of slow motion.
I enhance my prompts by integrating an LLM to automatically break motion into detailed, time-based descriptions.
If there is interest, I can share a workflow that integrates Ollama for this purpose.
Feel free to leave a comment and follow my social media if you’d like to see it released.
Reference & Credits
This valuable insight is based on the following post:
“WAN 2.2 Faster Motion with Prompting”
https://www.reddit.com/r/StableDiffusion/comments/1p54o54/wan_22_faster_motion_with_prompting_part_1/
Many thanks to the author for sharing this information.
Notes
This workflow is configured for the GGUF version, but fp8 and Distill versions can also be used by switching the model loader.
Designed with 480p output in mind.
There are many variations of Lightx2v LoRA, and their suitability depends on the type of video you want to generate.
If artifacts appear in 3-pass mode, try lowering the CFG value from 2.5.
Acknowledgements
Special thanks to Kijai for providing the excellent LoRA models and custom nodes.
Description
First Release Version