RepoPicks Radar

Updated every day around 23:30 UTC

Guides

ViMax

HKUDS/ViMaxReviewed by AI

Checked by RepoPicks Radar on

What is ViMax?

ViMax is a research project that chains AI steps to make a complete video from a sentence, a screenplay or a novel chapter: it writes the story, breaks it into scenes and shots, generates reference pictures, then generates and joins the clips. You drive it from the terminal, a text menu or a local web page. It brings no AI models of its own: you paste in keys for a chat service, an image service and a video service and pay those by the shot. The README lists Linux and Windows only, and you install it with Python and uv.

How easy is it to install ViMax?

D · terminal, several tools or a specific graphics card · Windows

  • Windows D
  • Linux D

There is no installer for either system: you install uv, clone the repository, run `uv sync` (it pulls PyTorch from the NVIDIA CUDA 12.8 index on Linux and Windows), copy configs/agent.example.yaml and hand-edit three blocks of model names, base URLs and API keys before anything runs. The browser interface needs Node.js 18+ and `npm install` on top of that. The README's Environment section lists only Linux and Windows, so there is no grade for macOS.

Coming soon: install in 1 click with the RepoPicks app.

ViMax on GitHub

Stars
12,888on Nov 10, 2026
Downloads
—all releases
Growth
+33stars in 6 days

Stars over time

12,85012,86012,87012,88012,890 Nov 4Nov 6Nov 8Nov 10 12,85012,86012,87012,88012,890 Nov 4Nov 7Nov 10

Facts

Runs on
Windows, Linux
You need to install
Python, NVIDIA graphics card
To build from source
Python, NVIDIA graphics card
Disk space
960 MB–4.3 GB · estimate
Accounts or keys
A chat/LLM service (OpenAI-compatible, for example OpenRouter or MiniMax), An image generation service (for example Google Nano Banana or GPT Image 2), A video generation service (for example Google Veo or Seedance)
Cost
Free, API costs extra · The code is free (MIT). Every shot is generated by a paid service (Google Veo, Nano Banana, Seedance, GPT Image 2 through OpenRouter and similar), and a one-minute video needs many of them.
License
MIT (Permissive)
Can replace
Runway, AI, LTX Studio
Last change on GitHub
Sep 30, 2026

ViMax risk signals (6)

  • Attentioncan place real people in generated video

    readme.md key features: "AutoCameo — place a person or pet from a reference photo into generated stories while maintaining a consistent appearance", and "Make Your Own Cameo Video ... by uploading your photo". The same feature works with a photo of someone who did not agree to it.

    AI review

  • Attentionpaid api cost grows fast

    readme.md Usage: "ViMax generates 2 keyframe candidates in parallel and selects one using your vision-capable chat model. This can improve visual quality and consistency, at a higher API cost." Nothing runs locally, so the chat, image and video services are billed for every scene, every candidate and every shot.

    AI review

  • Infoapi keys pasted into config files

    readme.md tells you to write `api_key: <YOUR_API_KEY>` into configs/agent.local.yaml, configs/idea2video.yaml and configs/script2video.yaml. Environment variables (VIMAX_LLM_API_KEY and friends) are offered as an alternative, but the documented path leaves three keys in plain text inside the project folder.

    AI review

  • Inforesearch code edited by hand

    readme.md asks you to type your idea, requirements and style directly into main_idea2video.py and the script into main_script2video.py, rather than passing them as input. The project is from a university group (homepage is the arXiv paper 2606.07649) and has a single GitHub release.

    AI review

  • Infoweb ui is a development server

    readme.md Web UI: `cd web && npm install && npm run dev`, then open http://127.0.0.1:4173. It is the development server, with no login described; the README tells you to reach a remote machine through an SSH tunnel instead of opening the port.

    AI review

  • Infomac not supported

    readme.md Quick Start, Environment block: "OS: Linux, Windows". macOS is not listed, and pyproject.toml only points torch at the CUDA index for linux and win32.

    AI review

Evidence is shown in its original language.

What Radar could not check about ViMax

  • How much a finished minute of video really costs at today's provider prices
  • Whether the final videos look like the demo clips in the README
  • Real disk use after dependencies, reference images and rendered shots pile up
  • Whether it runs at all without an NVIDIA graphics card
  • Whether the Web UI and the text interface are stable, since there is only one release

Similar projects

The RepoPicks app

Coming soon: install in 1 click with the RepoPicks app.

Before you download

See the release page on GitHub