Lance: video generator — what it does and what you need to run it

Lance is an open-source video generator from Bytedance Inc. It is an any-to-any model that, as described by the authors, is designed for understanding, generating, and editing images and videos. The code is published on GitHub, and the weights are available on Hugging Face.
What it does
The authors describe Lance as a native unified multimodal model with 3 billion active parameters. The stated areas of work are image understanding, video understanding, image generation, video generation, and image editing. In the repository, the model is tagged with image-editing, image-generation, image-understanding, unified-multimodal-models, video-generation, and video-understanding.
The export format specified by the authors is obj. This is a developer statement, not the result of independent measurement.
What you need to run it
Running it requires a CUDA platform. The authors require Python version 3.10 or newer and CUDA 12.4 or newer. As described by the developer, inference requires a GPU with at least 40 GB of VRAM — verbatim: "A GPU with at least 40GB VRAM is required for inference".
The largest model weights file takes up 13.2 GB, and all model weights files together take up 28.6 GB. The code is written in Python. The code license is Apache-2.0, and the model weights license is apache-2.0.
The repository was created on May 15, 2026, and the last code change was on July 14, 2026.
Who it is for
The model suits those looking for an open multimodal system for tasks involving images and videos, and who have access to hardware with CUDA and at least 40 GB of VRAM. This is a fairly high requirement, so running it on consumer GPUs with less memory is not expected, as described by the authors.
For those working on platforms without CUDA or with limited VRAM, Lance is not suitable. It is also worth noting that the obj export format is a statement by the authors, and before choosing a tool for a task, it makes sense to verify how well it matches your expectations.
Lance is an open model with code and weights under the Apache-2.0 license, focused on working with images and videos. The VRAM and CUDA platform requirements make it a tool for workstations with appropriate hardware.
Fact sheet
Repository · Model on HuggingFace · Developer’s site
| Task | any-to-any source |
|---|---|
| Code license | Apache-2.0 source |
| Weights license | apache-2.0 source |
| Platform | CUDA per the author’s description source |
| VRAM | 40 GB per the author’s description source |
| Largest weights file | 13.2 GB source |
| All weights files | 28.6 GB source |
| Export formats | obj per the author’s description source |
| Python | 3.10 per the author’s description source |
| Library | Lance source |
| Language | Python source |
| Last code change | 2026-07-14 source |
| Repository created | 2026-05-15 source |
| GitHub stars | 1338 source |
|---|---|
| Forks | 94 source |
| Downloads per month | 405 source |
| Model updated | 2026-05-28 source |
Values are collected automatically from official sources and were checked on 2026-09-08. Each one links to its source, and values taken from the developer’s pages also carry a verbatim quote — hover over the note. Pricing and versions are shown as of the check date and change most often; verify on the vendor’s site before buying.



