Wan 2.2 LoRAs (and Wan 2.1): keep your video LoRAs in one place
Wan 2.2 is Alibaba's open video family: a 5B text-and-image-to-video model (720p, 24 fps, for 8 to 12 GB cards) and 14B text-to-video and image-to-video models that use two experts, high-noise and low-noise (27B total, 14B active). Every Wan model shares the umt5-xxl text encoder. Video LoRAs are big, and each one is made for one exact base.
Keep every Wan LoRA, labelled by base, in your private vault.
Plans from 100 GB to 5 TB. From $4.99 a month.
Wan 2.2 vs Wan 2.1: which base your LoRA needs
For Wan video LoRAs, the model page names the exact base (Wan 2.2 14B high-noise, for example). Install only that one, and label each file with its base.
Wan 2.2 5B
The 5B model does text and image to video at 720p and 24 fps on 8 to 12 GB cards. Use LoRAs made for 5B.
Wan 2.2 14B high-noise / low-noise
The 14B models use two experts. A LoRA made for the high-noise expert goes with it, and the same for low-noise.
Wan 2.1 LoRAs
Many trending Wan LoRAs are made for Wan 2.1. Check the base named on the model page before you use one; we don't promise cross-use.
How big are Wan LoRAs?
In a sample of 50 Wan Video 14B text-to-video LoRAs on Civitai the median file was 293 MB, 98% were under 1 GB and the largest was 1,182 MB: about four times a typical Flux LoRA. They are .safetensors files.
| Plan | Storage | Price | About how many fit (293 MB median) |
|---|---|---|---|
| Locker | 100 GB | $4.99 a month | ≈ 341 |
| Library | 500 GB | $14.99 a month | ≈ 1,706 |
| Archive | 1 TB | $24.99 a month | ≈ 3,412 |
| Studio | 5 TB | $99 a month | ≈ 17,064 |
Where Wan LoRAs go, and how to keep them organized
ComfyUI and Forge folders
ComfyUI reads LoRAs from ComfyUI/models/loras/, Forge from webui/models/Lora/. One subfolder per family or base keeps things findable: loras/wan22-5b/, loras/wan22-14b-high/, loras/wan22-14b-low/, loras/wan21/. Then refresh the LoRA list.
Name files and keep trigger words
Name each file author-concept-version and keep its trigger words next to it, so you still know how to use it months later (how to install a LoRA, step by step).
Keep the whole folder in your private vault
Drag your whole models folder into BigVault. It sits in our own encrypted storage, and you can download it again on any computer (download scripts for Mac, Linux and Windows).
Add a Wan LoRA from Civitai in one click
Paste a Civitai model link in your vault and BigVault adds it in one click. It keeps the model's details and trigger words with the file, and tidies file names and descriptions. Imports run on our servers in the background, so you can close the page.
Building a collection? Read how to keep your own Civitai archive.
Render Wan video on Graydient (Beta)
Graydient has Wan 2.2 workflows, such as animate-wan22. Rendering a model from your vault on Graydient is in Beta and needs a Graydient plan; .safetensors files under 25 GB are compatible, apart from a few edge cases. Pasting vault links into the Graydient dashboard is coming soon.
Graydient: “Use 100,000+ Civitai models, instantly installable in one click.”
Wan LoRA questions
- Do Wan 2.1 LoRAs work on Wan 2.2?
- Check the base named on the LoRA's model page and use it with that base. We don't promise that a LoRA works across versions.
- Why are Wan LoRAs bigger than image LoRAs?
- Video LoRAs carry more weights. In a sample of 50 Wan 14B LoRAs on Civitai the median was 293 MB, about four times a typical Flux LoRA.
- Which Wan 2.2 model does my LoRA need?
- The one its model page names: Wan 2.2 5B, 14B high-noise or 14B low-noise, or Wan 2.1.
- Where do Wan LoRAs go in ComfyUI?
- In ComfyUI/models/loras/. One subfolder per base, such as wan22-14b-high, keeps them apart.
- Can I render Wan LoRAs from my vault?
- Graydient has Wan 2.2 workflows. Rendering a model from your vault on Graydient is in Beta, needs a Graydient plan and takes .safetensors files under 25 GB.
Video LoRAs fill a disk fast.
Plans from 100 GB to 5 TB. From $4.99 a month.
Questions? info@modeltrends.ai