--- title: "How to run Wan2.1 Locally | Open Source 🆓FREE Model on Your Own PC" source_url: https://youtu.be/RjrrJaoEMFk captured_at: 2025-10-19 author: "Prompt Engineer 48" contributor: "ป้าเกษตร" platform: youtube --- # How to run Wan2.1 Locally | Open Source 🆓FREE Model on Your Own PC youtube video by **Prompt Engineer 48**, 8m07s, uploaded 2025-03-03. URL: https://youtu.be/RjrrJaoEMFk Tags: #AI, #VideoGeneration, #Wan2.1, #Alibaba, #TongyiLab, #TextToVideo, #ImageToVideo, #OpenSourceAI, #DeepLearning, #MachineLearning, #AIVideo, #VideoEditing, #AIInnovation, #TechRevolution, #ArtificialIntelligence ## Shared in LINE Note+ - 2025-10-19 09:57 by ป้าเกษตร — note: บ้านบ้าน ## Description In this video, we are going to see the amazing capability of Wan 2.1, a flagship model from Alibaba Group. We will set the model up in our local system. We will see the result when ran in a Cloud GPU as well. Requirement: atleast 8 GB GPU Setup 1. Install Git: 2. Install dotnet: 3. Install SwarmUI on Windows: 4. Download Wan 2.1 model from Huggingface: Infact, download wan2.1_t2v_1.3B_bf16.safetensors and wan2.1_t2v_1.3B_fp16.safetensors, both and place it in the diffusion_models in the models folder 5. Download Clip: 6. Download VAE: 7. Download WorkFlow: Links: Blog: Github: HuggingFace: ------------------------------------------------ GPUs Used: Try Out Gloud GPUs on Novita AI (Affiliate Link): Get $20 to test out DeepSeek-R1: Try out RunPods (Copy the Entire Affiliate Link): ------------------------------------------------- CHANNEL LINKS: 🕵️‍♀️ Join my Patreon for keeping up with the updates: ☕ Buy me a coffee: 📞 Get on a Call with me at Calendly: 💀 GitHub Profile: 🔖 Twitter Profile: Time Stamp 0:00 Intro 0:47 Details 0:53 Examples 2:25 Installations 4:33 Running the Model 5:25 Starting the WorkFlow 6:41 Another Option of Using Hastags: #AI, ## Transcript (captions:auto:en-orig) introducing van 2.1 which is a groundbreaking open-source video Foundation model that redefines what possible in AI driven video creation this is created by the Toni lab of Alibaba group as you can see it can generate videos based on text images and other controlled signals and the one 2.1 series models are fully open source and we are going to test them out as well so we can see we have four models text two video 14 billion image to video two versions of 720p and 480p then text to video 1.3 billion now this 1.3 billion is you can use in your own local system if you just have AGP GPU that's what we are going to see in our video but before that let's see some of the examples and the wonderful capabilities of this model so we have this dog cutting vegetables I mean it's amazing flying dream looks pretty awesome this is the tilt shift and you can see it looks pretty good we have the whole universe inside the washing machine Eiffel Tower lights up we have some sort of in painting that you can do you can change the animal here of course there is out painting you can create the other musicians in the video we have this posture change and you can see that the entire environment is changed we have this black crow looks pretty awesome we have the monster family here just jumping around we have underwater scenery uh I mean scene it's pretty good archery I loved it love the fireflies at the back dirt race and it's amazing it's amazing mhm two cats fighting red and the yellow we have a French or a European dance here we have dog cycling okay pretty good looks awesome and there are other such examples that you can see here we are going to test everything out here so first of all you need to install git so go ahead and download git for Windows then you need to installnet 8.0 as well for Windows you need to go to swam UI and then download this thing we have the installer for Windows you can just download the installer keep it in a folder where there is space in that drive I'm going to put it um on the D drive I'm going to run this but there's some issue I already have this folder so I need to delete the swam UI folder next let's run this again so this will clone the repo and let you get started with the installation so cling the repo and starting the installation here so we have some steps we need to say next next and this will be downloaded so it's downloading everything there are a huge file about 6 GB files but we need to download that let me put it this in the shortcut if I need to to start it up so I can just double click and start this up so I need to download some models as well so I'm going to paste everything so I need this model to download and keep it in an appropriate folder so the folder that I'm going to need is download and I'm going to put it in the models and diffusion models here and the next thing that I need is so it's downloading I need this clip file I'm going to go ahead and download this clip file just a minute so download and put it inside the clip I'm going to need to download the vae files as well I'm going to paste in all the links don't worry so this is the vae file I'm going to go ahead and put it in the location next we are going to import some confy UI Json templates that you can use to run the models here so we waiting for downloading of the models and the download is complete now let's go use the system so if you remember we have this launch window button here the shortcut I double click and this is going to start up the Local Host 7801 once we are in the Local Host we click on generate and let's check the model that are going to use so we going to use when 2.1 click on vae we have the VA in place then in the models we're going to choose fp16 if you don't have this model you can download and keep it in the models folder next we are going to make this on text to video so we're going to put 25 frames and this video format is something you can choose uh I prefer to go with h265 MP4 having done this we go to confy work workflow and here we are going to put in the Json file that you have downloaded so if you remember text to video Json I just go and put it here so we have this just on so for example this is the positive prompt here this is the negative prompt here this is the Laten space so this is a space this is a dimension in which your diffusion models will be available or video will be available and ultimately you're going to see the output here so this is the workflow you can just click q and this will start up the system so I have a few working examples or outputs that you can see so this is one example given a positive prompt of film quality professional quality the video begins to show the surface of a pond and the camera slowly zooms into the closeup the water surface begins to bubble and then a blonde woman is seen coming off up the Lotus Pond and you can see the output here and that's the output okay another one is this a cute anime girl with massive ears and big tail and you can see that this is output here so I'm going to test out more examples in the future but this is how to get the system set up soort of you can go ahead and try and play with this system now there is another option of using that as well if you have Cloud GPU then you can directly go ahead so if I go to GitHub and you can see that we can CD clone or you can clone the report here change the directory and install all the requirements then download the using this commands here and you can just start it up using this command python generate pi and it can uh put the taso t2v you can use this model 1.3 billion if you are low on memory this is the size of the model this is the checkpoint this is the prompt so we can put two anamorphic cats uh boxing so I've tested that out as well on run pods and these are some of the output puts that you can see here so these are uh the outputs of two cats fighting with each other this is this was the lower version of the model this is another output and you can see that cybernetic Samurai glowing neon tatos and this was the another one we can see a grand medieval tournaments so these are some of the examples that I've tried and you can go ahead and test this model and let me know what do you think this model is able to do but that's all I got to talk about when AI this is interesting and I will see you in the next video bye