RAM, the Realtime AI Mixer, is a program that mixes AI-generated images in real time. It is built in vvvv gamma and uses StreamDiffusion, an open-source method that makes AI images fast enough for live use (StreamDiffusion). It comes as an app plus a separate AI Engine program that makes the images.
For us: RAM is the studio’s AI product, next to SpaceMusic. Version 1 exists, but sales, licensing and work on it are paused. The latest build is 0.7 from August 2025. The libraries under it are by Tebjan Halm.
Start here
- Latest build: Dropbox:
JHH SERVER/1_PROJEKTE/132_Realtime AI Mixer/05_RAM_RELEASES, folder250815_RAM EXE 0.7 - Manual (may be out of date)
- Working copy on the AI PC:
C:\AI
How to
Install RAM on a PC
- Copy the folder
250815_RAM EXE 0.7from05_RAM_RELEASES. The PC needs an NVIDIA graphics card. - Run its install/update batch for VL.StreamDiffusion.
- Run the model download batch. It loads the AI models from Hugging Face and needs disk space and time.
- Copy the Freedom.Redmond model into the HiResModels folder by hand.
- Start
RealtimeAIMixer.exe. The first start prepares the AI engine for this graphics card, which takes a long time.
Update the AI PC for development
- In
C:\AI, runInstallOrUpdateVLPythonNET.bat, thenInstallOrUpdateVLStreamDiffusion.bat. - Run
C:\AI\VL.StreamDiffusion\installer.bat, thendownload_models.batin the same folder. - To test, run
C:\AI\_Start_StreamDiffusion.bat. It opens the StreamDiffusion help patch in vvvv.
Make a new build
- Export the app from vvvv.
- Run
C:\AI\AfterExport.bat.
What we have
| Part | What | Where |
|---|---|---|
| RAM app | RealtimeAIMixer.exe, the mixer you start. | Builds in 05_RAM_RELEASES |
| AI Engine | The separate program that makes the images. | Next to the app |
| VL.StreamDiffusion | Tebjan’s vvvv library for live AI images, with the Python software under it. | repo, C:\AI\VL.StreamDiffusion |
| VL.PythonNET | Tebjan’s vvvv library that runs Python, RAM’s link to the AI models. | repo, C:\AI\VL.PythonNET |
| AI models | SD-Turbo, SDXL-Turbo, TAESD, ControlNets and Freedom.Redmond, from Hugging Face. | Downloaded by script into C:\AI |
Find it
- Dropbox:
JHH SERVER/1_PROJEKTE/132_Realtime AI Mixer(builds in05_RAM_RELEASES, studies in04_RAM_STUDIES, architecture diagrams in03_RAM_LAYOUT) - Notion: RAM (road to V1), Versions, Flatbook RAM, Potential issues, About diffusion models, HiRes models, RAM Tasks
- Sales one-pager: Dropbox:
JHH SERVER/1_PROJEKTE/132_Realtime AI Mixer/06_RAM_DOCUMENTS/251130_RAM_Pagelet - Update videos: Dropbox:
JHH SERVER/5_BIBLIOTHEK/22_Video Tutorials/RAM Updates - Repos: RAM-WebUI, DiffusionExplorer