I made an open-source tool called OSR6 Realtime Screen TCode.
It reads a selected region of your computer screen in real time and outputs low-latency TCode for OSR / OSR6 devices. The goal is simple: instead of preparing a script file first, you can watch something on screen and let the software generate live motion output.
Demo videos
Please do not repost these videos!!!
Here are a few adult demo clips showing the realtime screen-reading effect in action:
If there are any copyright concerns, please contact aivnailedeng@gmail.com and I will remove the videos.
Main features
Real-time screen-region analysis
Low-latency TCode output
USB serial and BLE connection options
Stable L0 stroke mode
Experimental six-axis mode for OSR6
Adjustable upper/lower limits for each axis
Measurement mode for saving safe personal limits
Sensitivity presets
Motion smoothing, rate limiting, travel scaling, and other tuning options
Script curve preview
English / Chinese interface
Windows build available, no installation required
Important note
Six-axis mode works best only when the lighting is good, the subject is clear, and the screen content is stable. For first-time use, I strongly recommend starting with L0 mode, using measurement mode to save comfortable upper/lower limits, and then trying six-axis only after everything feels stable.
This project was vibe-coded with AI assistance. I’m still improving the algorithms and UX, so feedback, issues, and contributions are very welcome.
Thank you for your support and interest! I’m glad to hear you are considering this for games without EDI mods.
To answer your question: I will definitely be adapting UDP support in upcoming versions, which should perfectly fit your wireless SR6 setup in the other room.
If you are in a rush and don’t want to wait for the official release, the source code is fully available on GitHub along with AI prompting guides. You can easily use an AI tool to generate a custom build with UDP support right now!
Also, I have just released version 1.1.2, which introduces GPU acceleration and specialized optimization for dance movements/animations. Feel free to check it out!
Just an idea have you considered adding a watch ahead mode for videos where the analysis would watch the video ahead of time of you on a separate window/monitor?
I feel like that might help with any real time de-sync for videos at least. Also where is the AI prompting guides on the GitHub?
Thanks for checking it out and for the suggestion!
Right now, the latency is already fairly low in both Hybrid Analysis and RTM Pose 2D with GPU acceleration, so I’m not sure a watch-ahead mode would improve the result enough to justify the extra complexity.
It would probably require a separate video-file or video-player workflow, because the current app is mainly designed around realtime screen-region capture. So it may be quite a lot of work, and the improvement might not be that big in practice.
But since the project is open source, feel free to download it and experiment with this idea yourself. If you manage to make a good watch-ahead version, I’d be happy to take a look.