Screen Recording Failed to Save: How to Recover What's Left
A recording that crashed, froze, or quit mid-session is rarely a total loss. How to tell what survived, rescue the audio, and stop it happening again.
You recorded for 35 minutes. You hit stop. Instead of a thumbnail, you get an error — or the app quits, or the machine reboots, or the video appears in your library but refuses to open. The work is apparently gone.
Usually it isn’t, or at least not all of it. A screen recording is not written as a single atomic file that either exists or doesn’t; it’s a folder of parallel streams that fail independently. Knowing which ones survive a crash, and which single missing piece makes an otherwise complete video unplayable, is the difference between a lost afternoon and a twenty-minute salvage job.
Why an Interrupted Recording Breaks the Way It Does
An MP4 file has a header-ish index called the moov box. It stores the frame table: where every frame lives in the file, how long it runs, what the timing is. Without it, a player has a long stretch of valid video data and no map to read it with, so it reports a corrupt file and gives up.
Here’s the catch: encoders write the frame data (mdat) continuously as you record, but they can only write the moov box at the end, once the final frame count and durations are known. Recorded’s encoder writes it on finalize and then relocates it to the front of the file for fast playback. So a recording killed mid-session leaves a large, perfectly good mdat with no index — a 2 GB file that no player will touch.
This is why file size is a misleading signal. “My recording is 1.8 GB but won’t open” is the normal shape of this failure, not a contradiction.
Step 1: Check What the Library Already Knows
Open your library before you go digging in Finder or Explorer. Recorded checks every session folder on scan and marks the broken ones for you.
A session whose save didn’t complete gets a Save failed badge and a line of text explaining what went wrong. The Play button is disabled on those cards and clicking the card won’t open the editor — deliberately, because a half-written file is exactly what makes an editor hang forever trying to parse it. Rename, Open Folder, and Delete all still work.
There are two distinct detections behind that badge:
- The save reported an error. When the save pipeline fails, Recorded writes a small
error.jsonmarker inside the session folder recording the message, the phase that failed, and a timestamp — and deliberately does not delete the folder, because whatever did get written is the only copy you have. - The save finished but the video was never finalized. Older builds could carry on past a failed finalize and leave an MP4 with no
moovbox and no error marker. Recorded now recognises those files by their shape and lists them as failed, so a file that will never play is labelled as such instead of silently hanging the editor.
In both cases the folder is kept, not cleaned up. That’s the point: the failure marker exists so you can decide what to rescue before anything gets deleted.
Step 2: Open the Folder and See What Survived
Use the Open Folder action on the failed card, or navigate there yourself. Recordings live under your user folder:
- macOS:
~/Movies/Recorded/Recordings/ - Windows:
%USERPROFILE%\Videos\Recorded\Recordings\
Each session is its own subfolder. Inside, the streams are stored separately rather than muxed together — which is what makes non-destructive editing possible, and also what makes partial recovery possible:
| File | What it is | Survives a crash? |
|---|---|---|
recording.mp4 | Screen capture, H.264 | Data yes, index often no |
mic_audio.wav | Microphone, uncompressed PCM | Usually yes |
audio.wav | System audio, uncompressed PCM | Usually yes |
webcam.mp4 | Camera, encoded separately | Same risk as the screen file |
metadata.msgpack | Timing, sync offsets, cursor data | Only if the save got that far |
error.json | The failure marker | Present when the save errored |
The important row is the WAV files. Because they are raw PCM written progressively rather than a container that needs a trailer, audio is the most crash-resistant thing in the folder. If you recorded a 40-minute narrated walkthrough and the video index is gone, you very often still have 40 clean minutes of your voice. For a tutorial or a meeting where the words were the substance, re-recording the screen against surviving narration is a far smaller job than redoing the whole thing. Open those WAVs directly in any audio app or editor — nothing special is needed to read them.
If metadata.msgpack is missing, the session was interrupted before the save completed, and the editor has no sync or cursor data to work with. Treat that folder as raw material, not as a project.
Step 3: Decide Between Repair and Re-record
Be honest about the economics here. Third-party tools exist that attempt to rebuild an MP4 index by scanning the raw stream and borrowing structure from a known-good reference file recorded with the same settings. They sometimes work. They also need a reference file, take real time to get right, and give you back footage with no guaranteed audio sync.
A useful rule: if the lost take is under ten minutes and easy to stage again, re-record it. If it’s a long session you cannot reproduce — a live meeting, a one-time interview, a bug that appeared once — the recovery attempt is worth the hour. Keep the original folder untouched and work on copies; there is no second chance at the source data.
Either way, don’t delete the failed session until you’ve decided. Delete is one click away on the card and it is not reversible.
What Actually Causes These Failures
In rough order of how often they bite:
The disk filled mid-session. This is the most common cause and the most preventable. Recording writes continuously and fast, and uncompressed WAV audio alone runs roughly 690 MB per hour for a stereo 48 kHz track. When the volume fills, the encoder can’t write its trailer and you get exactly the no-index failure above. Our disk space and storage guide has the arithmetic for estimating a long session before you start one.
The machine slept or the session locked. A display going dark is harmless. The system suspending is not — capture stops, and whether the file closes cleanly is a coin flip. Before a long unattended recording, set the sleep timeout well past your planned duration and make sure the machine is on mains power.
The app or the system was force-quit. Command-Q during a save, End Task, a kernel panic, a power cut. There’s no graceful path here; the finalize step simply never runs.
The output folder was a cloud-synced directory. This one is sneakier. On Windows, OneDrive’s Files On-Demand can offload a finished recording to the cloud and leave a placeholder behind; Recorded detects those placeholders and skips reading them during library scans, specifically to avoid downloading gigabytes of video on every refresh. But sync clients actively rewriting a file that an encoder is still appending to is a genuine corruption risk. If your Videos or Movies folder is inside a synced location, pause sync while you record.
Sustained encoder overload. Dropped frames alone don’t usually break a file, but an encoder that can’t keep up on a heavily loaded machine is in a worse position to recover from anything else going wrong. The performance optimization guide covers reducing that load.
Preventing the Next One
Most of this is habit rather than configuration:
- Check free space before long sessions. Not “is there some space” — is there two to three times what you estimate the session will write, including the audio.
- Record to a local internal drive. External USB drives, network shares, and sync folders all add failure modes to a process that writes continuously for an hour.
- Record in takes. The single best protection, and it improves the video too. A session broken into five-minute segments means a crash costs you five minutes, not fifty. Recording in takes covers how to work this way without the result feeling stitched together.
- Stop the recording before you do anything else. Don’t quit the app, close the lid, or switch users while a save is finalizing. The progress indicator is not decorative — it’s the trailer being written.
- Do a 30-second test first. Record, stop, play it back. A test catches a misconfigured device or a full disk before it costs you the real take, which is also point one on the pre-recording checklist.
- Export the finished video promptly. An exported MP4 is a finalized, self-contained file. A session folder sitting in the library is still a collection of parts. Export turns the fragile thing into the durable thing.
When the File Plays but Something Else Is Wrong
Not every bad recording is a failed save. If your file opens and plays but the picture is black, the audio is missing, or playback stutters, you’re looking at a capture or device problem rather than a file integrity problem — different diagnosis, different fixes. The troubleshooting guide walks through those in triage order, and organizing your recording library helps you find the right session quickly when you have dozens of them.
The Short Version
A crashed recording usually leaves most of its data intact, and the thing that makes it unplayable is a missing index, not missing footage. Check the library for the failure badge, open the session folder, and rescue the WAV audio first — it’s the piece most likely to have survived and often the piece that took the most effort. Then make the call between a repair attempt and a re-record based on whether the take is reproducible.
Then set up so it doesn’t happen again: local drive, free space, sleep disabled, and takes short enough that losing one doesn’t hurt. If you’re starting from scratch, download Recorded and get the test take out of the way before you need it.