Video generation
Make video on your Mac.
Text-to-video and image-to-video, generated entirely on-device. No cloud, no credits, no queue. As far as we know, no other local Mac AI app does this at all.
What you can do
- Text-to-video — describe a shot, render it locally
- Image-to-video — bring a still you already have into motion
- Seeds and frame counts — reproducible runs you can tune
- A local gallery — clips and their settings stay on your Mac
How it behaves
Video is heavy, so Bedrova is careful with it.
- Runs as a background job you can watch, cancel and come back to
- Counted in the same cross-model memory budget as everything else
- Refuses to start rather than pushing your Mac into swap
- Only licence-cleared pipelines ship; the rest are excluded on purpose
See the video docs for models, settings and limits.
From your own code
Video generation is an asynchronous job on the local API.
POST localhost:1337/api/v1/video/generations
{ "prompt": "a rover dog trotting through tall grass, warm afternoon light",
"mode": "text_to_video" }
GET localhost:1337/api/v1/video/generations/<id> Submit, poll, collect. Long renders survive app restarts, and cancelling frees the memory straight away.
FAQ
Can a Mac really generate video locally?
Yes. Bedrova runs a local text-to-video and image-to-video pipeline through MLX on Apple Silicon. Clips render on your Mac and land in a local gallery. Nothing is uploaded, and there is no queue or credit balance.
What do I need to run video?
Video is the most demanding thing Bedrova does. Plan on 32 GB of unified memory or more. The fit-check will tell you where you stand before anything loads, and the compatibility page breaks it down by RAM tier.
How long are the clips?
Short. Local video generation produces clips measured in seconds, not minutes, and render time scales with resolution and frame count. This is the same trade-off every local video pipeline makes today.
Can I drive it from my own code?
Yes. Video generation is exposed on the local API as an asynchronous job: you submit a request, poll for progress, and collect the finished clip. See the developer docs.
Which video models are supported?
Bedrova ships a reviewed, licence-cleared video pipeline rather than accepting arbitrary video repositories. Models whose licences do not permit redistribution are deliberately not shipped, curated or recommended, and Bedrova says so rather than quietly failing.