### 🔥 Key Features:
- 4 Permanent Power Loras (always active for consistent high quality)
-Power Loras (Fantom & Pusa included) — can be disabled if desired. Workflow still runs smoothly with only a small drop in quality.
- Auto Category LoRA/LoHa System — automatically applies the latest and best LoRA/LoHa based on the selected category while the 4 Power Loras stay constantly enabled
- Only the category-specific LoRA changes, making it fully applied to the video with zero hassle
- Image to Video support
- Sage Attention (toggle on/off)
- Final Upscale (default: OFF for maximum speed)
- Clean and stable output
### ⚡️ Performance:
- 8-second video (16 frames @ 16fps)
- First generation: ~8 minutes
- Subsequent generations: ~6 minutes
- Tested on RTX 5090
### How to Load:
Simply drag and drop the JSON file into ComfyUI — the workflow will load automatically.
### Model:
- Main: Wan2.2 14B Q8.gguf
- Compatible with Wan2.1 and Wan2.2
### Why This Workflow Stands Out:
This is the first workflow I built from scratch. After trying many different workflows over the past few months, none have matched this one in
### Model:
- Main: Wan2.2 14B Q8.gguf
- Compatible with Wa The Power Loras remain always on, while the category LoRA switches seamlessly and applies perfectly to the entire video.
After months of testing and refining, I'm excited to release my personal daily driver video workflow. This is the best workflow I've used so far in terms of quality, simplicity, speed, and reliability.
I will continue updating it in the future with better motion quality, new Loras, and more options.
### Requirements:
- ComfyUI (latest version)
- Wan2.2 14B Q8.gguf
- Recommended: RTX 4090 or higher (best performance on 5090)
### Tips:
- Enable Sage Attention for higher quality
- Keep Upscale disabled if you want maximum speed
- Keep your category Loras updated for best results
---
If you find this workflow useful, please like and follow! Drop your feedback and suggestions in the comments so I can make it even better.
Enjoy creating! 🔥
⚠️ Important Disclaimer:
All preview images and videos in this page are AI-generated and fictional. They do not represent real people or events.
By downloading and using this workflow, you agree that you are fully responsible for how you use it and the content you generate.
Description
First public release of my personal Wan2.2 14B Ultra Fast NSFW Video Workflow.
Includes:
- 4 Permanent Power Loras ( Fantom & Pusa )
- Smart Auto Category LoRA System
- Image to Video + Sage Attention + Upscale
- Very fast & stable generation
FAQ
Comments (9)
nice bro can you make armpits licking lora hehe i am just asking
Which pusa is it? I found several with the exact name but always a different hash (i.e. high).
Hi,
Pusa LoRA for Wan2.2 (High Rank - ~4.57 GB versions)
Here are the main official sources for Pusa LoRA:
Best & Official Source (Hugging Face) - Recommended
Repository: RaphaelLiu/Pusa-Wan2.2-V1
You can download high_noise_pusa.safetensors and low_noise_pusa.safetensors directly from there (the high-rank versions are large, around 4+ GB each).
Civitai Version (Easier for ComfyUI users)
Model Page: Pusa - Wan Video LoRA
Look for the High Rank 256 versions (these are the big ones ~4.57 GB).
Hi, thanks for the links. I found the same two and they do not share the same hash. That's why I was unsure which one to take (in case one is fundamentally different from the other). Thank you :)
Pusa isn't implemented for native....so like any other lora it will do -something- but not as it's intended. if you're using KJ's wrapper here (+ Euler or Flowmatch_Pusa) then it will have an effect)
"### ⚡️ Performance:
- 8-second video (16 frames @ 16fps)
- First generation: ~8 minutes
- Subsequent generations: ~6 minutes
- Tested on RTX 5090"
I have no idea how you can make RTX 5090 that "slow with Wan 2.2.
### Tips:
- Enable Sage Attention for higher quality (slightly slower)"
You've got it exactly wrong. SageAttention makes generation FASTER but slightly sacrifices quality. What is this nonsense?
Hey!
Thanks for the feedback! 😊
I just shared my personal experience and mentioned that this is my first time building my own personal workflow.
If you have some time and feel like it, I’d really appreciate it if you could point out any issues and help me fix them.
I’m here to learn new stuff too!
Thanks a lot!
Guess I'm missing something. My video ends up looking like a fogged up window.