Wan2.2-TI2V-5B

Wan2.2-TI2V-5B

Generates 720P video at 24 fps from text, or from an image plus text, and runs on a single consumer graphics card such as the 4090. A 5B video model from Wan-AI for creators and researchers.

Made by Wan-AI on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Wan2.2-TI2V-5B is the 5B text-and-image-to-video model in the Wan2.2 family from Wan-AI. One model turns a text prompt, or an image together with a prompt, into a video at 720P and 24 frames per second.

It uses the Wan2.2-VAE, which compresses video by a factor of 16x16x4. The authors say this lets it run on one consumer graphics card such as the 4090.

Who it suits: Video creators, researchers and developers who want an open video generator that fits on a single high-end gaming GPU instead of a server.

Industry | Who uses it | How they use it | Example
Media | Video creators | Turn a text prompt into a short clip | Generate a 720P scene from a description
Marketing | Content teams | Animate a still image with a prompt | Bring a product picture to life
Research | Video generation researchers | Study a compact open video model | Compare it with the larger A14B models
Software | Developers of creative tools | Call it through Diffusers or ComfyUI | Add video generation to a workflow

Skip it if you have no GPU at hand, since video generation needs graphics hardware.

Where to use it:

• ComfyUI, where the card marks the 5B integration as done
• Diffusers, which the card also marks as done for this model
• The command line tools in the Wan2.2 GitHub project
• A single consumer GPU such as the 4090
• Windows, macOS or Linux with Python, to download the files

Use case | Who | What they get back
Text to video | Creators | A 720P, 24 fps clip from a prompt
Image to video | Designers and marketers | A clip that starts from your picture and follows your prompt
Research comparisons | Researchers | A fast 5B model to test against bigger ones

How to set it up:

• Python and pip
• Git, to copy the Wan2.2 repository from GitHub
• PyTorch 2.4.0 or newer
• A consumer GPU such as the 4090

1. Clone the Wan2.2 repository from GitHub and open its folder.
2. Install the dependencies from its requirements file.
3. Download the TI2V-5B model files.
4. Follow the repository instructions to generate a video.

# Ensure torch >= 2.4.0
pip install -r requirements.txt

pip install -U huggingface_hub
hf download Wan-AI/Wan2.2-TI2V-5B

How to use it: Give the model a text prompt, or an image and a prompt, and ask for a 720P, 24 fps video. The visible part of the card shows no generation command, so use the repository instructions.

Example: You describe a scene in a sentence and run the generation script from the repository. The model returns a short 720P video at 24 frames per second matching the description.

For newcomers and experts: If you are new to video models, begin with a short text prompt and a plain scene, then try adding an image to see how it starts the clip.

If you know the field, the Wan2.2-VAE compresses by 16x16x4. The Wan2.2 family also introduces a mixture-of-experts design into video diffusion, with specialized experts across denoising timesteps, and trains on curated data labeled for lighting, composition and color tone.

Text prompt, or image plus prompt
Wan2.2-VAE latent space at 16x16x4 compression
Diffusion model denoises the video
720P, 24 fps video

Good to know:

• The card calls it one of the fastest 720P at 24 fps models available, and that is the authors' claim
• Hugging Face lists 16 quantized and 137 fine-tuned versions made by other people
• Wan2.2 was trained on more data than Wan2.1

Title: Extra training data in Wan2.2 compared with Wan2.1
More images: 65.6 percent
More videos: 83.2 percent

• License: Apache 2.0
• Cost to run: none named, apart from the consumer GPU the card mentions
• Data it reaches: the prompts and images you supply, plus the files you download
• Not for: the source states no restrictions

On Hugging Face, 13 November 2026

• Task: text to video
• Library: wan2.2
• Formats: Safetensors
• Downloads in the last 30 days: 18,923
• Likes: 973
• License: apache-2.0
• Languages: en, zh
• Quantized versions: 16
• Fine-tunes: 140
• Tags: wan2.2, diffusers, safetensors, ti2v, text-to-video

Links

• Model: https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B
• Files: https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download Wan-AI/Wan2.2-TI2V-5B

Then follow the usage in its model card. Nothing in this kit runs by itself.