Post
998
LatentObserver PRO
obsxrver
AI & ML interests
Researcher
Recent Activity
liked a Space about 16 hours ago
obsxrver/wan2.2-i2v-lora-demo liked a model about 21 hours ago
deepseek-ai/DeepSeek-V4-Pro updated a Space about 22 hours ago
obsxrver/wan2.2-i2v-lora-demoOrganizations
Posts 2
Post
5305
(https://github.com/obsxrver/wan22-lora-training)
If you’ve been wanting to train your own Wan 2.2 Video LoRAs but are intimidated by the hardware requirements, parameter tweaking insanity, or the installation nightmare—I built a solution that handles it all for you.
This is currently the easiest, fastest, and cheapest way to get a high-quality training run done.
Why this method?
* Zero Setup: No installing Python, CUDA, or hunting for dependencies. You launch a pre-built [Vast.AI](http://Vast.AI) template, and it's ready in minutes.
* Full WebUI: Drag-and-drop your videos/images, edit captions, and click "Start." No terminal commands required.
* Extremely Cheap: You can rent a dual RTX 5090 node, train a full LoRA in 2-3 hours, and auto-shutdown. Total cost is usually $3 or less.
* Auto-Save: It automatically uploads your finished LoRA to your Cloud Storage (Google Drive/S3/Dropbox) and kills the instance so you don't pay for a second longer than necessary.
How it works:
1. Click the Vast.AIÂ template link (in the repo).
2. Open the WebUI in your browser.
3. Upload your dataset and press Train.
4. Come back in an hour to find your LoRA in your Google Drive.
It supports both Text-to-Video and Image-to-Video, and optimizes for dual-GPU setups (training High/Low noise simultaneously) to cut training time in half.
Repo + Template Link:
https://github.com/obsxrver/wan22-lora-training
Let me know
if you have questions
If you’ve been wanting to train your own Wan 2.2 Video LoRAs but are intimidated by the hardware requirements, parameter tweaking insanity, or the installation nightmare—I built a solution that handles it all for you.
This is currently the easiest, fastest, and cheapest way to get a high-quality training run done.
Why this method?
* Zero Setup: No installing Python, CUDA, or hunting for dependencies. You launch a pre-built [Vast.AI](http://Vast.AI) template, and it's ready in minutes.
* Full WebUI: Drag-and-drop your videos/images, edit captions, and click "Start." No terminal commands required.
* Extremely Cheap: You can rent a dual RTX 5090 node, train a full LoRA in 2-3 hours, and auto-shutdown. Total cost is usually $3 or less.
* Auto-Save: It automatically uploads your finished LoRA to your Cloud Storage (Google Drive/S3/Dropbox) and kills the instance so you don't pay for a second longer than necessary.
How it works:
1. Click the Vast.AIÂ template link (in the repo).
2. Open the WebUI in your browser.
3. Upload your dataset and press Train.
4. Come back in an hour to find your LoRA in your Google Drive.
It supports both Text-to-Video and Image-to-Video, and optimizes for dual-GPU setups (training High/Low noise simultaneously) to cut training time in half.
Repo + Template Link:
https://github.com/obsxrver/wan22-lora-training
Let me know
if you have questions
models 9
obsxrver/wan22-aot
Updated
obsxrver/wan2.2-i2v-blink-piss
Image-to-Video • Updated • 2.86k • • 14
obsxrver/wan2.2-t2v-bdsm
Text-to-Video • Updated • 130 • 8
obsxrver/wan2.2-i2v-lightx2v-260412
Updated • 19
obsxrver/sam3
Mask Generation • 0.9B • Updated • 772
obsxrver/wan2.2-t2v-scat
Text-to-Video • Updated • 649 • 25
obsxrver/wan2.2-t2v-piss
Updated • 48 • 5
obsxrver/wan2.2-i2v-scat
Image-to-Video • Updated • 933 • • 17
obsxrver/Wan2.2_GTA-Style
Text-to-Video • Updated • 21 • 2
datasets 0
None public yet