A video export can have the expected duration and still cut a frame early. To make that mistake visible, I used two synthetic clips with a source label burned into each frame. I wanted to check whether a small Shotcut project would keep its edit after a full restart.
Put a label on both sides of the cut
Each clip was 150 frames at 25 fps. I kept source frames A025 through A124, then appended all of B. For an inclusive trim, the first segment is 124 − 25 + 1 = 100 frames. That makes the join output frames 99 and 100, even though the corresponding source labels are A124 and B000.
| Output frame (zero-based) | Expected source label |
|---|---|
| 0 | A025 |
| 99 | A124 |
| 100 | B000 |
| 249 | B149 |
Check the two frames around the join. Checking only the first and last frames can miss a boundary mistake.
Verify the saved edit and the output separately
I saved the native MLT file, exported, quit Shotcut, reopened the file and exported again. Both saved projects recorded the same two playlist entries. Both exports had 250 decoded frames, 1280 × 720 resolution, 25 fps and no audio stream. FFmpeg decoded both files without error, and the four sampled labels matched the table in each export.
A first pass at trimming raced the seek operation and landed on the wrong range. Waiting for the seek to finish, then checking the saved playlist before exporting, caught it. I also had to select the export controls again after restarting. Keep a record of the export settings separately from the edit.
The Shotcut test notes and side-by-side frame captures show the setup and result. The test used Shotcut 26.11.15 on Windows and one silent synthetic project. It did not cover Premiere import, missing-media relinking, effects, audio or hardware encoding.
Disclosure: I maintain FOSS Desk. AI assisted with drafting this article from the recorded test evidence.
