AInVFX/ainvfx-paintaplate

AInVFX PaintAPlate · test plates

The plates of AInVFX PaintAPlate, a ComfyUI node pack and page: you paint on the first frame of a shot to erase, add or relight, and get the shot back in the plate's own format. These 23 shots are the ones the app is tested and demonstrated on.

Every plate is AI-generated, made in ComfyUI. LTX 2.5 (Lightricks) made each shot and its sound track, from a first frame by MOSS-Image (OpenMOSS) or FLUX 3 Image Klein (Black Forest Labs), or from text alone. No real people, no real locations, no stock footage. Please say the plates are AI-generated when you show them.

Use them with PaintAPlate

On a Runpod pod: PLATES=1 bash tools/runpod_start.sh, or bash tools/get_plates.sh once ComfyUI runs. Either puts the plates in ComfyUI's input/plates folder, and the Open plate dialog lists them with a thumbnail. On your own machine:

hf download AInVFX/ainvfx-paintaplate --repo-type dataset --include "*.mp4" --local-dir ComfyUI/input/plates

The plates

All 23 have 121 frames at 24 fps (about 5 seconds): H.264 video with an AAC sound track, in MP4. Each file carries one metadata comment, an AI-generation label that names the models. The ComfyUI generation graphs were removed from the files; the video and sound streams are unchanged.

File	Size	Frames	fps	Weight
1280p_car.mp4	1280 x 704	121	24	2.2 MB
1280p_packshot.mp4	1280 x 704	121	24	1.8 MB
1280p_truck.mp4	1280 x 704	121	24	5.3 MB
boat.mp4	1920 x 1080	121	24	19.2 MB
boat_2d.mp4	1920 x 1080	121	24	12.5 MB
chalet_2d.mp4	1920 x 1080	121	24	6.0 MB
chalet_v2.mp4	1920 x 1080	121	24	6.9 MB
gs_car_front.mp4	1920 x 1080	121	24	3.1 MB
gs_car_side.mp4	1920 x 1080	121	24	4.1 MB
horse.mp4	1920 x 1080	121	24	6.9 MB
lighthouse.mp4	1920 x 1080	121	24	6.8 MB
lighthouse_2d_b.mp4	1920 x 1080	121	24	6.0 MB
rooftop.mp4	1920 x 1080	121	24	5.3 MB
set_planet.mp4	1920 x 1080	121	24	5.7 MB
set_station.mp4	1920 x 1080	121	24	3.8 MB
spaceship.mp4	1920 x 1088	121	24	5.9 MB
square_2d.mp4	1920 x 1080	121	24	10.5 MB
square_v3a.mp4	1920 x 1080	121	24	6.9 MB
tea.mp4	1920 x 1080	121	24	4.4 MB
topdown.mp4	1920 x 1080	121	24	6.2 MB
train_2d_v3.mp4	1920 x 1080	121	24	9.2 MB
train_v2.mp4	1920 x 1080	121	24	4.8 MB
umbrella.mp4	1920 x 1088	121	24	4.6 MB

License

This dataset: CC BY 4.0. Use, share and change the plates, with credit to AInVFX.

Notes on the licenses of the models that made them, read on 27 to 29 October 2026:

• LTX 2.5 is under the LTX-2.x Community License. Lightricks claims no rights in its outputs (section 5). The same license counts a model trained on LTX outputs as a derivative of LTX-2.x (section 3.2): read it before you train on these plates.
• FLUX 3 Image Klein is under the FLUX Kommunity License. Its section 4(b) does not allow using its outputs to improve another AI model with a similar function.
• MOSS-Image is under Apache 2.0.

This section is a summary, not legal advice. It follows the AInVFX statement of AI ethics: documented data, labelled generated content.

Who made it

Made by AInVFX. We teach generative AI for film and TV in the Generative AI Bootcamp for Film and TV: 60 hours of live training over several weeks, online and in English, built around your own project, with cloud GPU included. It runs several times a year: see the upcoming cohorts on ainvfx.com/bootcamp.

Questions, results, bugs: contact AInVFX or open a discussion on this repo.