Different r_frame_rate, avg_frame_rate, and time_base values do not by themselves mean a video is damaged. Record all three for each relevant stream: the two frame-rate fields are reported rate values, while time_base is the stream’s time-unit scale. None alone proves the actual on-screen motion cadence. If that cadence matters, examine frame timestamps and review representative playback.
Start with the stream, not a single number
Use ffprobe’s stream report and identify the video stream by its index, codec_type, and, where useful, disposition or language. A file may contain multiple video streams, such as a main picture and an attached cover image; do not assume that every stream-level value describes the picture you intend to deliver. Audio and subtitle streams have their own timing information too, but they are not the video frame rate.
The ffprobe documentation describes -show_streams as reporting information for each media stream. Its JSON writer is useful when you need to preserve the exact fields for a delivery note. Record missing or unavailable values as such; do not silently substitute a value from another stream.
Read the three fields as different evidence
r_frame_rate is libavformat’s estimated base rate, not a measurement of each frame’s spacing. avg_frame_rate reports an average. time_base supplies timestamp units in seconds: 1/90000 means one ninety-thousandth of a second per tick. Preserve their separate labels and ratios. FFmpeg’s AVStream reference documents these roles and warns that the base-rate estimate is a guess.
For a diagnostic record, do not relabel the time unit as fps or treat differing rate fields as proof of damage. When cadence matters, gather timestamp and playback observations rather than choosing whichever metadata number matches the delivery requirement.
Make a per-stream record in order
- Run
ffprobe -v error -show_streams -of json "INPUT_FILE", replacingINPUT_FILEwith the path to your media file. This command is an instruction, not a recorded test result. - Locate the intended video stream. Note its stream index and confirm
codec_typeisvideo; distinguish it from any attached picture or other video stream. - Copy
time_base,r_frame_rate, andavg_frame_rateexactly as reported. Preserve each ratio; do not round fractional rates to integers. - Add the file’s intended delivery requirement beside these observations, if one exists. A requirement such as “24 fps” is a separate acceptance criterion, not a reason to reinterpret the other fields.
- Mark cadence as not established by these fields unless you have checked timestamps or representative playback. Keep that status separate from whether the metadata report itself can be read.
The -select_streams and -show_entries options can narrow reports to particular streams and fields; the documentation explains their stream-selection and section syntax. For a first inspection, a full stream report is often less error-prone than guessing the correct stream number in a filter.
Example: differing values, no damage verdict
The following is a hypothetical report for one intended video stream, not a measured file:
| Stream | time_base | r_frame_rate | avg_frame_rate | What to conclude now |
|---|---|---|---|---|
| index 0, video | 1/90000 | 30000/1001 | 24000/1001 | Record the mismatch; cadence and file integrity remain unverified |
This row does not prove that playback alternates between two rates, that one field is “wrong,” or that frames are missing. Nor does it show that the file is healthy. It tells you what ffprobe reported for that stream and which question remains open. Compare the observations with the delivery contract, but do not fail the file solely because the two rate fields differ.
When cadence matters, inspect timestamps and playback
If an editor reports uneven motion, a delivery specification requires a particular cadence, or the rate-field mismatch affects a conversion decision, move to evidence closer to playback. ffprobe can report packet and decoded-frame information with -show_packets and -show_frames; the documentation distinguishes packet reports from frame reports. As an author-recommended diagnostic check, inspect presentation timestamps over a representative interval and note their spacing, gaps, or repeated values. This is not a vendor-certified damage test. Seeking and selected intervals can be imprecise, so a short excerpt is not necessarily a complete-file conclusion.
Then play representative sections at normal speed in a media player or editor. Check both a typical passage and any section where motion looks uneven. Timestamps describe timing metadata; visual playback helps answer how motion is presented. Neither a single field nor a tiny sample proves every part of a long file behaves identically. Record the stream, interval, timestamp observations, and playback result before deciding whether a conversion or further investigation is needed.
FAQ: Which rate should I put on a delivery sheet? Use the rate required by the delivery specification, then report the stream’s ffprobe values separately. If the specification asks for actual playback cadence, verify it with timestamp inspection and playback rather than choosing one field by name.
FAQ: Does `time_base: 1/90000` mean 90,000 fps? No. It describes timestamp units. The denominator is not a frame-rate reading.
FAQ: Do different `r_frame_rate` and `avg_frame_rate` values prove variable frame rate? No. They are a reason to investigate when cadence matters, not a standalone proof of variable cadence, damage, or correct playback.
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.