I build local AI studios that run on your own GPU, with no cloud and no account: Remiqora for music and Remiqora Video for video. Both are open source, and both install only the models' official weights.
A local AI music studio. ACE-Step 1.5, YuE2-3B and Stable Audio 3 in one interface: describe a song or a sound
in your own words and a local language model writes the prompt; split the track into stems, add parts with
the AI arranger and finish it in the built-in multitrack editor.
Windows / Linux (NVIDIA) · macOS (Apple Silicon) Website · GitHub · Download 
|
A local AI video studio. A couple of sentences become a storyboard written by a local language model;
MiniMax H3, LTX 2.5 / 2.3 or WAN 2.2 shoot it scene by scene with sound. Then montage with a DaVinci Resolve
export, a node graph for versions, music videos cut to a song and a 1080p / 4K / HDR10 finish.
Windows · NVIDIA RTX, 16 GB Website · GitHub · Download · Wiki 
|
Licenses. The apps and the models they run are licensed separately:
- Remiqora: MIT. Its models: ACE-Step 1.5 MIT; YuE2-3B weights CC BY-NC 4.0 (non-commercial); Stable Audio 3 weights Stability AI Community License (commercial use is free under US $1M annual revenue, after a free registration with Stability AI); Qwen3.5 (the prompt assistant) Apache-2.0.
- Remiqora Video: GPL-3.0. Its models: MiniMax H3 Community License; LTX-2 Community License; WAN 2.2 Apache-2.0; Krea 2 Community License. A video falls under the license of the model that made it.
Also contributed: Remiqora in the audio.cpp projects list.
If my work is useful to you, you can support it on Ko-fi.
Feedback welcome.