The repair asks for a healthy reference video. You have plenty of clips, but which one belongs with the damaged recording? Start with an original from the same recording setup, then check the video and audio settings before using it.
What the reference supplies
A reference helps a repair tool interpret media that survived in a damaged MP4 or MOV. It supplies a working example of the recording structure and codec configuration. It does not donate replacement scenes or sound to the damaged file. If part of the event was never written, a healthy clip cannot supply that missing event.
This requirement is specific to the recovery method. For example, untrunc’s usage instructions call for a working video, ideally from the same camera. A file that already has readable streams may have another route; see the readable-file checks in our moov guide before assuming every damaged video needs a reference.
Look beside the damaged recording first
A healthy clip recorded immediately before or after the damaged one is a useful candidate if the settings stayed the same. Use the file straight from the recording device or app. A file sent through an editor, messaging app or social platform may have been converted. Matching its filename extension or visual appearance tells you little about the recording setup.
For a phone, the camera app matters as well as the phone model. For a camera, check the recording mode rather than just its make. For OBS, use the same recording profile and output setup. A screen capture and a camera recording can both be MP4 files while carrying quite different streams.
Find more than one plausible candidate if settings changed around the interruption, and label what you know about each. Keep guesses visible: “same camera, frame rate unknown” is more useful than silently treating an uncertain clip as a match.
Match the streams, including audio
Compare the following against the original recording settings or a known healthy clip from that session. The damaged file may no longer expose these details.
- Video codec and recording mode. H.264 and HEVC are different codecs; choosing the same resolution does not make them interchangeable. Check high-efficiency, slow-motion and other special modes rather than assuming the camera’s default was used.
- Resolution and frame rate. Match the recorded dimensions and intended frame rate. A 1920×1080 recording is different from a 3840×2160 recording. If you know the original used 29.97 fps, do not round that to 30 when choosing a candidate.
- Audio codec, sample rate and channels. Keep the audio setup consistent, including mono versus stereo. A silent scene can still have an audio stream; a recording made with audio disabled may have none.
- Number of recorded audio tracks. An OBS clip with one mixed track is different from a recording with separate microphone and desktop tracks. OBS’s multiple-track guide shows that both source routing and the enabled recording tracks determine what is saved.
These checks narrow the candidates; they are not a compatibility certificate. MendVid’s recovery routing uses the reference’s codec and audio layout, and some paths infer timing from it. A wrong reference can therefore affect the result beyond whether the file opens.
Read a candidate without changing it
If FFmpeg is installed, run its companion ffprobe on the healthy candidate:
ffprobe -v error -show_entries stream=index,codec_type,codec_name,profile,width,height,r_frame_rate,avg_frame_rate,sample_rate,channels -of json "healthy-reference.mp4"The command reads the file and prints stream information. Repeat it for another healthy candidate and compare the video codec, dimensions and audio entries. FFmpeg documents the selection syntax in its ffprobe reference. It does not repair the recording or verify that every frame decodes.
In the output, 30000/1001 is approximately 29.97 fps. Treat frame-rate fields as clues, especially for variable-frame-rate recordings; an average alone does not describe the original timing. If the damaged MP4 returns moov atom not found, its missing stream report is not evidence that it had no audio.
As an illustrative comparison, suppose a session used H.264 video and two AAC audio streams. A candidate showing HEVC video and one AAC stream does not match that known setup, even if its picture looks similar. When the original settings are unknown, ffprobe can describe a candidate, but cannot prove it matches the damaged file.
If you need to record a new sample
Use the original device and app where possible, restore the known recording settings, and stop the new recording normally. The sample can show an ordinary scene; reproducing the lost event is unnecessary. Include sound if the original recorded sound, and check that the sample itself opens and plays.
Do not record the sample onto the only card holding the damaged recording. Copy the existing files to separate storage and use another card for the sample. Keep the damaged original untouched throughout the attempt.
There is no universal sample length that guarantees a match. A short, normally finalized clip is a practical starting point; it still needs the right streams and settings. If the camera or app has since been updated, note that change. If you no longer have it, look for an untouched clip from the same session before treating an internet download as equivalent. Another device of the same model is a candidate to assess, not a confirmed substitute.
Use the reference, then check the result
In MendVid’s MP4 and MOV repair workflow, a required reference stays on the same paid job without another payment. New purchases were closed when checked on 1 November 2026. When available, the US$9.99 fee buys a repair attempt; confirmed payment comes before upload, repair and preview. An unsuccessful attempt is not refunded. The price and attempt terms explain that decision before checkout.
After a repair, check the full result’s opening, middle and ending, then listen to every expected audio track. Compare the duration with what you expected and watch for sound drifting away from the picture. A matching reference does not establish that the entire recording survived, and a short preview cannot establish that the ending is intact.
Keep the candidate reference, the settings you could confirm and the remaining uncertainties together. If the result has the wrong duration or missing sound, that record gives the next investigation somewhere concrete to start.