Brimkern video pipeline β BRIK (int8)
AnimateDiff-Lightning motion modules on an SD 1.5 base, converted to the BRIK format so textβvideo runs inside the browser on the visitor's GPU (WebGPU). No inference server, no upload: the prompt and the frames never leave the machine.
| file | size | role |
|---|---|---|
video-unet-q8.brik |
914 MB | SD 1.5 UNet (epiCRealism), int8 |
video-motion-1step-q8.brik |
483 MB | AnimateDiff-Lightning 1-step motion modules, int8 β the app's default from its next deploy |
video-motion-2step-q8.brik |
483 MB | AnimateDiff-Lightning 2-step motion modules, int8 (?vsteps=2) |
video-motion-q8.brik |
483 MB | AnimateDiff-Lightning 4-step motion modules, int8 (?vsteps=4) |
video-clip-q8.brik |
131 MB | CLIP-L text encoder, int8 |
rife-4.26-f16.safetensors |
11 MB | RIFE 4.26 frame interpolation, f16 |
A Lightning checkpoint fixes its own step count; only the motion module changes between them (same UNet, same CLIP). The app loads one of them: full pipeline 1.54 GB (UNet + one motion module
- CLIP + RIFE + the 4.7 MB TAESD decoder, fetched from its own repository).
Frame interpolation. AnimateDiff produces 16 unique frames (1.3 s at 12 fps). RIFE then inserts 3 frames between each pair (the last pair wraps to the first, since the clip loops): 64 frames, a 4-second clip at 16 fps. It smooths and slows the motion; it does not invent new motion. The port runs on the engine's WGSL kernels and matches the PyTorch reference to 5Β·10β»β΅ (max abs error on an interpolated frame).
ByteDance describes the 1-step checkpoint as provided "for research purposes". We made it the default because waiting time decides whether anyone uses the feature: on our checks (a beach at sunset, a cat walking in a garden, 16 frames, fixed seed) it renders the same scene and motion as the 2-step one, slightly softer.
Try it
π https://brimkern.com/chat β Browse / load a model β the video card.
This is a pipeline, not a single-file model: several files load together, so the ?model=repo
deeplink does not apply. Every file is streamed by HTTP Range, cached, and reusable offline.
β οΈ Honest cost, announced before the click: 1.54 GB to download, then about 10 s of GPU work per clip on an M4 Mac (measured, see below), longer on smaller GPUs. Clip length is adjustable from 16 to 24 generated frames (~4 to ~6 s once interpolated) (8 frames was removed: the motion modules are trained on 16, and below that the frames fall apart). A progress fraction (UNet blocks per step, then frames decoded) drives a real progress bar and a remaining-time estimate β a diffusion wait with no feedback is indistinguishable from a hang.
Measured
The video path shares the SD 1.5 UNet with the image pipeline, so it inherits its measured kernel
work: the tiled int8 3Γ3 convolution (Γ1.84 on that kernel, Γ1.67 end-to-end on a 256 px image).
Replayable benches in
scripts/e2e/.
End to end, one 16-frame 256 px clip β Mac M4 (24 GB), Chrome, production build, GPU duty 1, seed 42, same prompt, settings alternated, two runs each, two series on 2026-10-08 (the 1-step module ran in both and gave 10.0 s every time):
| motion module | seconds per clip |
|---|---|
| 1-step (default) | 10.0 / 10.0 / 10.0 / 10.0 |
| 2-step | 16.0 / 16.5 |
| 4-step | 28.1 / 28.1 |
Frame interpolation on top (same machine and settings, alternated): 16 frames without RIFE 8.5 / 8.5 s, with RIFE Γ4 (64 frames, 4 s) 10.5 / 10.0 s.
These include two engine changes of the same week, each measured against a switch that turns it off: convolutions as an implicit GEMM (8-frame clip 28 β 23 s) and a clip shown as soon as the GPU is done (the WebM encoding used to record a ~10 s loop in real time first: β9 to β10 s per clip). Other GPUs will differ; the benches are replayable.
Format
A .brik is a self-describing container: topology and quantization tiers travel inside the file,
and every shard is one contiguous HTTP range β which is what makes a 1.53 GB pipeline loadable in a
tab. Specification:
BRIK_FORMAT.md.
License
Derived weights; the sources keep their terms:
- AnimateDiff-Lightning (ByteDance/AnimateDiff-Lightning) β creativeml-openrail-m.
- epiCRealism (emilianJR/epiCRealism, SD 1.5 fine-tune) β creativeml-openrail-m.
- RIFE 4.26 (hzwer/Practical-RIFE, weights from hzwer/RIFE) β MIT, Β© 2021 hzwer; converted to f16 safetensors without other change.
- TAESD decoder (madebyollin/taesd) β MIT, not mirrored here.
The Brimkern engine itself is MIT. Converting weights does not change their license.
Model tree for romainkh14/brimkern-video-BRIK
Base model
ByteDance/AnimateDiff-Lightning