Analysis updated 2026-05-18
Generate a walking, dancing, or jumping animation for a VRM avatar just by typing a description.
Preview a generated character motion instantly in the browser before saving it.
Save generated animations as reusable .vrma files for use in other VRM-compatible apps.
Run motion generation fully offline and unlimited using the local ARDY engine instead of a paid API.
| kirakun0328/text-to-vrma | gaearon/gitbook-plugin-prism | gaearon/react-hot-api | |
|---|---|---|---|
| Stars | 88 | 88 | 88 |
| Language | JavaScript | JavaScript | JavaScript |
| Last pushed | — | 2022-11-06 | 2016-03-14 |
| Maintenance | — | Dormant | Dormant |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 1/5 | 3/5 |
| Audience | vibe coder | writer | developer |
Figures from each repo's GitHub metadata at analysis time.
The prebuilt Windows release needs no setup, running from source needs Node.js, and the optional ARDY engine needs Python and about 20GB of downloads.
Text-To-VRMA is a web app that turns a written description, such as walk in place or wave hand, into a motion file for VRM avatars, a common 3D character format. You type a sentence, the AI generates a matching animation, and the app builds a VRMA file right in your browser, playing it on a 3D character on screen so you can preview it before saving. There are two ways to generate the motion. The first uses an OpenAI API key or a Codex subscription to produce keyframes directly, which needs no extra setup and works on any computer, and is good at precise poses, hand gestures, and facial expressions. The second, newer option runs an NVIDIA research model called ARDY on your own computer to generate full-body movement, such as walking, dancing, or jumping, trained on real motion capture data, producing higher quality movement with no per-use cost and no internet connection required once it is set up. When both are available, the app can automatically split a request between the two: an AI model handles understanding the request and translating it into English, while ARDY generates the actual body movement. Generated animations also include facial expressions like smiling, surprise, or blinking, and you can choose whether to include those or only the body movement when saving the file. The interface supports Japanese, English, Chinese, and Korean, switching instantly from a menu. A prebuilt Windows download needs no setup at all. Running it from source offers a simple browser version, which needs only Node.js and a few commands, or a desktop version built with Electron that adds support for signing in with a Codex subscription and for setting up the ARDY engine from inside the app. The ARDY engine itself is optional, requires Python and about 20 gigabytes of downloads, and runs faster with an NVIDIA graphics card, though it can also run on a regular processor at reduced speed. The developer has only tested the app on Windows 11, though it should also work in other browsers since it is a web app. Generating motion with an API key costs a small amount of money per use, depending on the model and animation length.
Text-To-VRMA turns a written sentence into a 3D animation file for VRM avatars, previewed live in your browser.
Mainly JavaScript. The stack also includes JavaScript, Three.js, Electron.
No license information is stated in the README.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.