Spaces:

vclmax2
/

frame-compositor

Running on Zero

App Files Files Community

frame-compositor

Commit History

Jul 13

17:30
migrate layer schema from float seconds to integer frame numbers⁠ Running

vclmax2 +1: Claude Sonnet 4.6

852b4bb
16:21
fix frame timing: use round() to match JS Math.round for frame range calc

vclmax2

9aec4ee
15:59
fix audio: drop PyAV audio mux, use ffmpeg stream copy from source

vclmax2

cfc5a23
14:53
reduce ZeroGPU duration to 10s

vclmax2

bd0978f
14:44
revert to 7.2s build (pre-multiplied cache, pinned memory, threaded audio)

vclmax2

12387ea
14:41
optimize: fp16, torch.compile blend, double-buffered decode, batch=64, torch.no_grad

vclmax2

cd68686
14:36
optimize: pre-multiplied static composite cache, parallel layer render, pinned memory, threaded audio

vclmax2

66a6eb6
14:32
optimize: static UV cache, threaded encoder, batch=32, animated warm-start

vclmax2

b117472
14:25
full GPU pipeline: batched Newton solve + grid_sample on CUDA, precomputed pixel grid

vclmax2

19c2773
14:18
match JS bilinear quad warp (lerpQuad) instead of perspective transform

vclmax2

e66c1d6
14:04
replace kornia with cv2.warpPerspective — exact same logic as original render_editor_video.py

vclmax2

9db7a97
13:54
fix text wrap: use font.getbbox() to match original render_editor_video.py exactly

vclmax2

6daf007
13:43
fix homography: use get_perspective_transform(src,dst) + align_corners=True to match OpenCV exactly

vclmax2

11c92e8
13:26
Fix warp_perspective homography direction: dst→src for backward warp

vclmax2

c3fdb4a
13:11
Fix homography direction: src=layer corners, dst=frame corners

vclmax2

4245b07
13:07
Fix layer tensors not on DEVICE inside GPU fn, fix src_pts off-by-one

vclmax2

ab345c6
13:01
Fix frame normalization: divide by 255 after torchcodec decode

vclmax2

ac2f33c
12:56
libx264 ultrafast encode, remove PyNvVideoCodec

vclmax2

cc1992d
12:50
Add PyNvVideoCodec GPU encoder with libx264 fallback

vclmax2

127f4b1
11:53
Fix audio muxing: use codec name instead of unsupported template= kwarg

vclmax2

6b69b9d
11:49
Replace decord with torchcodec for native CUDA frame decode

vclmax2

ebe8ab8
11:38
Batched homography per layer + h264_nvenc encoder with libx264 fallback

vclmax2

3215f22
11:31
Add application file

vclmax2

b3b0e62
11:20
add frame-compositor space

vclmax2

Verified signature
3ae109d
11:15
initial commit

vclmax2

Verified signature
840b7cf

Commit History

Jul 13

migrate layer schema from float seconds to integer frame numbers⁠ Running

fix frame timing: use round() to match JS Math.round for frame range calc

fix audio: drop PyAV audio mux, use ffmpeg stream copy from source

reduce ZeroGPU duration to 10s

revert to 7.2s build (pre-multiplied cache, pinned memory, threaded audio)

optimize: fp16, torch.compile blend, double-buffered decode, batch=64, torch.no_grad

optimize: pre-multiplied static composite cache, parallel layer render, pinned memory, threaded audio

optimize: static UV cache, threaded encoder, batch=32, animated warm-start

full GPU pipeline: batched Newton solve + grid_sample on CUDA, precomputed pixel grid

match JS bilinear quad warp (lerpQuad) instead of perspective transform

replace kornia with cv2.warpPerspective — exact same logic as original render_editor_video.py

fix text wrap: use font.getbbox() to match original render_editor_video.py exactly

fix homography: use get_perspective_transform(src,dst) + align_corners=True to match OpenCV exactly

Fix warp_perspective homography direction: dst→src for backward warp

Fix homography direction: src=layer corners, dst=frame corners

Fix layer tensors not on DEVICE inside GPU fn, fix src_pts off-by-one

Fix frame normalization: divide by 255 after torchcodec decode

libx264 ultrafast encode, remove PyNvVideoCodec

Add PyNvVideoCodec GPU encoder with libx264 fallback

Fix audio muxing: use codec name instead of unsupported template= kwarg

Replace decord with torchcodec for native CUDA frame decode

Batched homography per layer + h264_nvenc encoder with libx264 fallback

Add application file

add frame-compositor space

initial commit