Check the file that will actually be sent. A preset named “1080 delivery” does not prove 1920 × 1080 pixels, the intended audio tracks or the correct timeline version. Metadata establishes structure; watching and listening establish whether the content is right.
On this page
1. Make a local report
With FFmpeg/ffprobe installed from a trusted distribution, run the command below against your own file. It reads the media and writes a JSON report to a separate file. The report may contain filenames and tags, so review it before sharing it publicly.
ffprobe -v error -show_format -show_streams -of json "delivery.mp4" > "delivery-report.json"Open Creator Edit Lab’s media report inspector and choose the JSON file or paste its contents. The page parses the report locally and does not upload it. Enter expected stored width/height, audio stream count or maximum decimal MB to turn a passive report into explicit comparisons.
2. Read the evidence correctly
| Field | Useful interpretation | Common mistake |
|---|---|---|
| format.size and format.duration | File bytes and reported duration; derive whole-file average rate | Calling the derived rate the video-only bitrate |
| width / height and rotation | Stored pixel dimensions plus orientation metadata | Treating a rotated portrait file as necessarily wrong |
| codec_name / profile / pix_fmt | A more specific description than the extension | Assuming every decoder accepts the combination |
| avg_frame_rate / r_frame_rate | Two reported rate fields, including fractions | Diagnosing VFR from their difference alone |
| audio streams / channels | Tracks and channel count are separate | Assuming two audio tracks mean stereo |
| color tags | Metadata supplied in the report | Assuming tags prove an appropriate color transform |
3. Use a mismatch as a reason to investigate
The inspector’s synthetic example has 1920 × 1080 video, one two-channel audio stream, a 60-second duration and 77,400,000 bytes. Require 3840 pixels of width and the comparison must report a mismatch. Require 1920 and it matches the stored width. Neither result proves the correct edit was exported. These values are a test fixture, not a measured camera file.
Missing fields are reported as unknown or not reported, not silently replaced with zero. A report with multiple video streams needs manual selection of the intended stream; brief comparisons use the first non-attached-picture video stream. An album-cover image is excluded from that selection when ffprobe identifies it as attached artwork.
4. Review what metadata cannot establish
- Watch the beginning, a demanding middle section and the end of the exact output file.
- Confirm the intended revision, titles, complete sequence range and any slate or handles.
- Listen to dialogue and each required audio track; check sync at several identifiable events.
- Review captions with the actual playback route and verify that burn-in is not doubled by a sidecar track.
- Check color in the intended workflow and seek to later sections instead of only playing from the start.
5. Keep an acceptance record
Download the inspector summary and keep it with the delivery brief and a separate human review note. The downloadable summary omits the source filename and arbitrary metadata tags. It records only the inspected fields, warnings and comparisons. A later revision needs a fresh report: a passed check on version 3 does not validate version 4.
Budget the upload with the delivery time estimator and examine captions with the Subtitle QA Desk.
Sources and verification
Documentation checked October 28, 2026. The worked numbers are illustrative calculations, not hardware benchmarks. Instructions describe a reproducible procedure; software labels can vary by release. How we verify and correct our work.
Try an actual generated-file report
The 240-frame fractional-rate case includes a real generated MP4 and its sanitized ffprobe report. Download the JSON, paste it into the inspector and require 320 × 180 pixels with zero audio streams. The comparison should match, with 24000/1001 fps and 10.01 seconds. Change the expected width to 1920 to observe a mismatch. This exercise tests the inspection workflow using a synthetic file, not a camera recording.
Build on what you just learned
Your next useful step.
Budget the size, inspect the export, keep a record.
Generative Extend: a few extra frames still need an editorial review
I would use generated extensions for a deliberate transition or timing need, then inspect the new frames and identify them in the project record.
Read the guide →Budget video bitrate for a hard file-size limit
Allocate the total size budget across duration, audio and a stated reserve before choosing video bitrate. Then measure the actual exported bytes, especially when…
Read the guide →
