Spaces:
Running on Zero
Running on Zero
Commit History
3 days ago
-
test: a dark black-and-white night and a soft snowfall scene get no note from the picture checkβ
Unverified signatureThe labelled set grew to 42 outputs (27 good). No good file fires a condition. Two of the new ones are hard negatives, now kept as fixtures with tests that they yield no note: - A low-key black-and-white night (true blacks, dim highlights, no colour). It passes two of veiled's three tests (p99 66.7 < 135, colourfulness 0.7 < 20) and is kept out by its blacks: p1 1.0 against 40. It is also now the worst good file for flat: luma std 14.9 against 6, 2.49x clear. - A cabin in heavy snowfall at dusk, soft by content. It fails all three veiled tests and sits 2.01x clear of static. veiled needs both ends of the range pulled toward grey, and no good file has both. The three good files with lifted blacks all keep highlights of at least 176.8 (the overcast beach, 1.31x above 135). The one good file with dim highlights has p1 1.0. The pinned numbers in the test now say that, and the provenance comment in outputs.py gives the 42-file counts. No threshold moves.0b00a5b -
fix: the picture check names flat, veiled and static pictures from measurements on real outputs, at one size, and tells the user as a hintβ
Unverified signatureThe old check gated on two ratios (8x8 structure ratio < 0.70 AND high-frequency ratio > 1.0) placed between two populations that were both garbage: colour static from the broken Space build and flat grey. On a labelled set of 38 real outputs (23 good, 3 flat grey, 9 veiled, 3 colour static, every one looked at), it let all of the colour static through. It said nothing about any veiled picture. Its ratios also changed with resolution: the operator's grey fails at 512x960 and passes once resized to 256 px. Its message quoted the garbage ranges as "real renders". assess_picture() resizes up to 16 evenly spaced frames to one analysis size (long side 256 px, area-averaged; numpy only), takes medians, and names at most one of three conditions: - flat (no picture): luma std < 6. Worst grey 4.3, worst good 22.0. - veiled (a picture under a grey veil, which pulls both ends of the tonal range toward grey): luma p1 >= 40 AND luma p99 < 135 AND Hasler-Susstrunk colourfulness < 20. Veiled frβ¦752b18e
4 days ago
-
test: reachability, ZeroGPU wiring and frame-structure runs carry their mode's inputsβ
Unverified signatureThese helpers called generate() for a mode with nothing it requires: a text mode with no prompt, image_to_video with only a text shot row, and audio_to_audio with a path to a file that does not exist (the loader was stubbed, so nothing ever opened it). generate() is about to refuse all three before it reserves a GPU, which would leave these specs measuring the refusal rather than the paths they are about. The run helpers now pass conftest.inputs_for(mode), the audio probe passes a real WAV, and the shot rows become one keyframe -- an image at frame 8, strength 0.8 -- which reaches conditioning.py by the same route. Green with and without the scoping change, in both environments.f80c73b
-
fix: motion is reported, never refused β the temporal gate demoted after it destroyed a real renderβ
Unverified signatureMIN_MEAN_FRAME_DELTA = 0.5 refused image_to_audio_video at 512x960x121: structure 0.9974, high-freq 0.0559, block std 81.6, luma band 238.0 β a coherent photographic scene, a red balloon over a lake at dawn β thrown away for mean |Ξ| 0.4071. It is near-static because it is conditioned on a still image of that scene, which is the mode working as designed. The gate never had the margin it appeared to have. Measured with the runner's own decoder on the post-int8-fix m2 runs, the only banked artifacts whose frames have been looked at and are photographs: 0.3859, 0.4071, 0.4168 β against the operator's colour static at 0.403. A legitimate render sits BELOW the failure the threshold was calibrated against. The two populations overlap, and no threshold on this axis separates them, because a temporal measure cannot tell a still scene from a dead one: both measure near zero, and what distinguishes them lives inside the frames rather than between them. So mean |Ξ| WARNS. It is still measured, β¦8c042d4 -
feat: measure whether there is a picture in the frames β structure, not distinctnessβ
Unverified signatureEvery content check this project had could be passed by structureless garbage, and one artifact proved it: 121 of 121 distinct frames, the geometry the request asked for, non-zero bytes, a full luma range β and it was colour static. space_matrix scored artifacts like it as PASS, and the app inspected the frames not at all (generate.py:597 checks only frames-non-empty; generate.py:690 and outputs.py check only that a path exists and is not 0 bytes). The reason no arrangement of those numbers could ever have worked is that distinctness, byte count, geometry and brightness range are properties of NOISE. Noise has all of them in abundance. What noise does not have is structure at a scale above one pixel. Two ratios say it, both divided by the frame's own luma std so they describe shape rather than contrast β a dim shot and a bright shot of the same scene land in the same place: SR = std(8x8 box means) / std(pixels) real 0.918..0.994, failure 0.530 HF = std(laplacian4) / std(pβ¦e54477c