Use FFmpeg’s signalstats to compare how luma values are distributed across frames, not to assign a quality grade. Read YAVG for average luma, YMIN and YMAX for extremes, and YLOW and YHIGH for the 10th and 90th percentiles. When a frame-to-frame comparison matters, add YDIF as a clue about the average difference between corresponding Y-plane samples. None of these values provides a universal pass threshold or diagnoses the cause of a change on its own.
Choose frames and read the measurements
For a creator checking whether a short edit contains a possible brightness jump, compare frames that represent the concern: one before it, one at it, and one after it. This is a comparison method, not a requirement imposed by FFmpeg.
- Identify the timecodes of the frames you want to compare. Keep the selected frames in the same sequence and note whether a cut, fade, or other intended transition occurs between them.
- Use
signalstatsmetadata for those frames. FFmpeg documents per-frame Y values in the 0–255 range and showsffprobeoutput for selectedYMINandYMAXtags. For other fields, request the correspondinglavfi.signalstatsframe tags using the same documented metadata pattern. See the signalstats filter documentation. - Record the five distribution readings for each frame. Compare like with like: average against average, lower percentile against lower percentile, and so on. Do not interpret a low
YMINalone as proof that the frame has crushed shadows, or a highYMAXas proof of a defect. - If you need to investigate a change between adjacent frames, record
YDIFalongside the distribution readings. FFmpeg defines it as the average difference between Y-plane samples in the current frame and corresponding samples in the previous input frame. The documented frame-metadata example can help orient you to the tag naming; the comparison and decision rules here are editorial guidance, not built-in FFmpeg verdicts.
Hypothetical comparison worksheet
The figures below are hypothetical diagnostic data, not results from a software test or official acceptance limits. In this example, frame B's average and upper percentile are higher, while its minimum and lower percentile are close to frame A's. That pattern is a reason to inspect what is brighter in the image, not proof that the whole frame is overexposed.
| Frame | YMIN | YLOW | YAVG | YHIGH | YMAX | Observation and next decision |
|---|---|---|---|---|---|---|
| A | 4 | 24 | 82 | 150 | 230 | Reference: record the frame content and whether its distribution matches the intended shot. |
| B | 5 | 25 | 105 | 190 | 235 | Average and upper percentile rise; check whether bright areas, framing, or a deliberate transition explain the shift. |
For a practical acceptance record, add the timecodes, whether the frames are adjacent, and the observed YDIF value where relevant. Decide whether to inspect further based on the edit’s intent and visible content. There is no general numeric cutoff in these readings that makes a frame acceptable or unacceptable.
Limits and what to do next
YDIF is a whole-plane average difference measure as documented; it does not identify which objects or regions changed. A larger reading alone cannot establish flicker, a bad cut, or an exposure problem. A smaller reading does not prove that a particular subject remained visually consistent. Motion, a scene change, and image changes can all be reasons to investigate the context rather than label the number.
If the readings shift unexpectedly, inspect the corresponding frames in sequence and compare the actual content around the timecodes. Check whether the change coincides with an edit, movement, or an intended lighting change. If the problem appears localized, look at that region in the image rather than expecting a frame-wide average to explain it. If the numbers and visual impression disagree, keep both observations: signalstats describes its measurements, while your review addresses the intended appearance. The filter output does not determine a correction or guarantee that detail can be recovered.
Common questions
Does a high `YMAX` mean highlights are clipped?
Not by itself. It reports a frame’s maximum Y value; the number alone does not establish whether important detail has been lost. Inspect the affected image and its context.
Should I compare `YAVG` or `YHIGH` first?
Start with the measurement tied to your question. Use YAVG for a shift in average luma and YHIGH when you want to compare the upper part of the distribution. Neither substitutes for reviewing the frame.
Can `YDIF` confirm flicker?
No. It measures average differences between corresponding Y-plane samples in neighboring input frames. Use it to locate changes for review, not as a standalone flicker detector.
Sources and next step
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the film workspace. The external software controls discussed above belong to their respective products.