Video Generation on a Laptop
Hello!
This workflow utilizes a few custom nodes from Kijai and other sources to ensure smooth performance on an RTX 3050 Laptop Edition with just 4GB of VRAM. It's optimized to improve generation length, visual quality, and overall functionality.
🧠 Workflow Info
This is several ComfyUI workflow capable of running:
2.0-ALL -- Includes all workflows:
Wan2.1 T2V
Wan2.1 I2V
Wan2.1 Vace
Wan2.1 First Frame Last Frame
Funcontrol (experimental)
Funcameraimage (experimental)
Coming soon: Inpainting experimentals get updated
🚀 Results (Performance)
*to be updated
🎥 Video Explainer (Vace edition):
🎥 Installation Guide (V1.8):
📦 DOWNLOAD SECTION
⚙️ Nodes Used (Install via ComfyUI Manager or links below)
Note: rgthree Only needed for Stack Lora Loader
📦 Model Downloads
*these are conversions from the original models to run on less VRAM.
most versions
Faster/Better quants for i2v
fun,inpainting,T2V,Vace
fun-control
🔗 WAN2.1 Fun-Camera-control 14B GGUF
fun-Camera-Control
All these GGUF conversions are done by:
https://huggingface.co/calcuis
https://huggingface.co/QuantStack
*If you cant find the model you are looking for check out there profiles!
🧩 Additional Required Files (Do not downlaod from Model Downloads)
📥 What to Download & How to Use It
✅ Quantization Tips:
Q_5 – 🔥 Best balance of speed and quality
Q_3_K_M – Fast and fairly accurate
Q_2_K – Usable, but with some quality loss
1.3B models – ⚡ Super fast, lower detail (good for testing)
14B models – 🎯 High quality, slower and VRAM-heavy
Reminder: Lower "Q" = faster and less VRAM, but lower quality
Higher "Q" = better quality, but more VRAM and slower speed
🧩 Model Types & What They Do
Wan Video – Generates video from a text prompt (Text-to-Video)
Wan VACE – Generates video from a single image (Image-to-Video)
Wan2.1 Fun Control – Adds control inputs like depth, pose, or edges for guided video generation
Wan2.1 Fun Camera – Simulates camera movements (zoom, pan, etc.) for dynamic video from static input
Wan2.1 Fun InP – Allows video inpainting (fix or edit specific regions in video frames)
First–Last Frame – Generates a video by interpolating between a start and end image
📂 File Placement Guide
All WAN model
.gguffiles →
Place them in yourComfyUI/models/diffusion_models/folder⚠️ Always check the model's download page for instructions —
Converted models often list exact folder structure or dependencies
🔗 Helpful Sources:
Installing Triton: https://www.patreon.com/posts/easy-guide-sage-124253103
Common Errors: https://civarchive.com/articles/17240
Reddit Threads:
https://www.reddit.com/r/StableDiffusion/comments/1j1r791/wan_21_comfyui_prompting_tips https://civarchive.com/articles/17240
https://www.reddit.com/r/comfyui/comments/1j1ieqd/going_to_do_a_detailed_wan_guide_post_including
🚀 Performance Tips
To improve speed further, use:
✅ Xformer
✅ Sage Attention
✅ Triton
✅ Adjust internal settings for optimization
If you have any questions or need help, feel free to reach out!
Hope this helps you generate realistic AI video with just a laptop 🙌
Description
FAQ
Comments (27)
Interesting. Sorry to hear about the process time :( but at least it worked!!!
no worries atleast I got it to work on my laptop! :)
@The_frizzy1 I just tried it on a rx 7900 xtx on the size you had on the workflow. It would take 20 mins. But I like to test before I fry my gpu, so I start small :D
@nerfme haha thats great 24gb is alot more than i had so 20 mins is pretty doable for me.
I would love to see some results on here if you got the time to post them below.
I have just used a flux workflow to create an image and then make it a video/Animted it.
@The_frizzy1 Oh it will be a while. I have a frenzied chat-gpt4-o in my ear trying to tease me out of getting a workflow for an issue I have. Just so it can go to town on images I have to make prompts.
@nerfme are you using zluda or directml for the generation
@kavirap_23 Zluda 3.8.6
@nerfme How is the performance of AMD graphics cards
@Cijade It works. Nvidia has better support, and works faster. If you're looking to buy a card for AI use, Nvidia wins.
whats model GGUF do you use? i only see Wan2_1-T2V-1_3B_bf16.safetensors..........3gb , because other model is 17gb and 15gb for FP8
I use wan2.1-t2v-14b-q8_0.gguf Yeah well its a video model the original is like 25gb+ there are some other gguf conversions like https://huggingface.co/city96/Wan2.1-T2V-14B-gguf/tree/main but I have found this works the best with my incredibly low vram
what is apply riflexrope please help?
Its frame generation so rather than making every frame some frames are interperlated using the other frames. This speeds it up uses less vram and allows you to make longer videos if you have enough vram you dont need to have this.
thank you for the reply im not sure where to add download or insert code could you help
@The_frizzy1 I join the question. I rarely download anything directly from github and I can't figure out how to upload it to ComfyUI. I can't use the console for some reason.
@faust_777 Im not really sure what you guys need.
To install from github you can download the zip or clone the repo directly into your comfy ui folder via cmd.
https://www.gitkraken.com/learn/git/github-download#how-to-clone-a-github-repository
2025.2.26 RIFLEx is supported in ComfyUI-HunyuanVideoWrapper by KiJai.
So just use the comfy UI manager and install the custom node.
@The_frizzy1 Thanks, now I've figured out how to download and install from github without a manager. RifleX installed, but the node still doesn't work (remains red). Maybe there's something else I don't understand...
@faust_777 in the workflow the node comes from KJ nodes if you downloaded it directly it’s gonna be under a different name look for it in the nodes and replace it.
@faust_777 Нужно изменить security level с normal на weak в файле confiq.ini в comfyui и тогда сможете с git устанавливать что угодно.
did you say any hunyuan lora are working with wan2.1?
Some are but I am still testing it so you’ll have to see. If you’re just doing text I have a workflow for hunyuan aswell, the image workflow is not great you can always use the hunyuan wrapper if you have enough vram. I am working to update this to speed it up with sageatn and triton I will try more loras before updating.
i use t2v and i wonder if i can run hunyuan loras on your workflow in wan 2.1 1_3b model
i was not expecting the 14B model to run on 4gb vram damn, but i just wanted to ask how long does it take to generated on rtx 3050 4gb as you said, but also the 8gb version if its any faster or ...
Yeah it runs but I have waited up to 5 hours for about 3 secounds... but now im just going with lower quantization and the new update finally has teacache and torchcompiling so its alot faster now.
I have to defenelty make a post to show how long what took and i only have the 3050 so thats all i can say.
Would love to see someone else though with more vram show there results!
I have not tested this, but with default settings with 16gb of VRAM using the Q5 quantized level (so like 13gb of vram), its 18 mins (or 40s per increment) at the most efficient, but 20-30 when it struggles with a motion. . This will happen more when it gets confused with a hand, it seems like anyway.
18 mins with T2V and 8 mins with I2V with 13gb. Its still slow, like twice as slow as Huanyun, but it has much better temporal understanding. Will be mindblowing with loras, Ive even setup my own, just dont want to wait the 12+ hours of training... lol
3050 with 8 GB VRAM + 32 GB RAM & wan lora takes me approx. 25 min. i2v including upscaling and interpolation for 3 sec
