Data Recovery Case File · Action Cameras · Present but Unplayable
The picture stops, the sound continues: reading a stream-level failure by its strangest symptom — and a trip's footage reassembled to run
This batch's other action-camera case met an emptied card; his arrived full — and unwatchable, which is the stranger and more instructive presentation. Back from a trip, his 1TB SanDisk Extreme Pro lists its MP4 files everywhere it's asked; some show thumbnails; but playback, on the camera and in an editing app alike, follows one script: a couple of seconds of footage, then the image freezes — "however the sound is still playing." The most recent files "simply don't play" at all. No visible damage to the card; contact by email, please (honoured throughout). A video that dies while its own audio marches on is not random corruption — it's a precise clue about where in the files' anatomy the damage lives, and the decode is the road home.
| Device | SanDisk Extreme Pro 1TB SD — action-camera trip footage; files listed with sizes and (partial) thumbnails intact |
| Reported playback | Consistent pattern across players: ~2 seconds of video, then frozen frame with continuing audio · newest recordings refuse to open entirely · no further recording since; email-only contact observed |
| Fault class | Card-level degradation striking the video streams' territory — the files' skeletons and light-bitrate audio surviving reads that the heavy video data fails |
| Equipment used | Write-blocked chip-level imaging (ACE Lab PC-3000 Flash) · stream-level container analysis and repair · fragment reassembly · full-duration playback verification |
The decode: why the audio outlives the picture — and what the two-second mark means
An MP4 is not one stream but a braid: video and audio recorded side by side, interleaved through the file in alternating chunks, with an index telling players where every chunk lives. The two strands are wildly unequal — video is the heavyweight, consuming the overwhelming bulk of every second's data; audio is a whisper riding alongside — and that asymmetry is the whole explanation for his symptom. When a card's memory degrades, failing reads land across its territory by area — so the dense video chunks, occupying almost all the ground, absorb almost all the damage, while the audio's slender track threads through comparatively unscathed. A player hitting the first unreadable video chunk does exactly what he watched it do: holds the last good frame (the freeze) while the intact audio strand plays on — the sound continuing is the braid's thin strand surviving the field the thick one couldn't cross. The consistent two-second mark dates the damage's onset within each file's layout; the thumbnails surviving (tiny, separately stored images) fit the same size-based logic; and the newest files refusing to open entirely points at the recordings written as the card's decline steepened — their opening structures themselves caught in failing territory. Read together, the symptoms acquit the files' structure (listings, sizes, indexes largely intact) and convict the card's media beneath the video chunks — which is genuinely the hopeful version: data that fails to read at consumer speed is a different prospect from data that was never written, and the bench's whole trade is reading the unreadable margin.
The recovery — the braid, re-woven
The card was imaged once, write-blocked at chip level on the PC-3000 Flash — the degraded regions under the video chunks negotiated in managed, repeated passes, marginal cells coaxed across the threshold consumer readers abandon at — and the footage rebuilt on the copy, file by file: video chunks recovered from the difficult territory and re-woven with their waiting audio; container indexes repaired where the newest files' openings had been caught; and each rebuilt recording verified the only way video can be — played, end to end, the frozen-frame script replaced by footage that runs. The honest ledger did what this genre's ledgers must: the substantial majority of the trip recovered to full playback; a bounded population of moments — chunks sitting on the card's worst cells — recovered with brief visible imperfections at those instants, noted per file; and a small remainder in the steepest-decline recordings itemised as beyond coherent rebuild. Delivery by email-arranged handover on new media, the card retired with its terabyte ambition and its actual condition both recorded — and the delivery note carrying the working-shooter's line this case earns: files that list but stutter are a card mid-decline; the trip after the first freeze is the trip you lose.
Outcome
The trip watchable again, imperfections named, losses counted — and the pattern filed for every shooter meeting the frozen frame. Video-freezes-audio-plays is a location report: the damage lives in the card's media beneath the heavyweight video data, not in your files' souls — stop playing the clips (every playback is a re-read of the failing territory), stop importing attempts, and absolutely stop recording on the card. Present-but-unplayable outranks empty as a prognosis: the structures survive; the reads are the battle, and managed chip-level reading wins most of it. And hear the card's chronology: newest-files-worst means decline was underway while you shot — the freeze pattern on one trip's early files is the warning for its later ones, and the retirement of a stuttering card is always cheaper than this page. His email asked for a quote to recover the footage. The answer, verified at full duration: watched.
Footage that freezes while the sound plays on
Stop playback attempts — each one re-reads the failing regions under your video — and retire the card from the camera immediately; no more recording, however much space remains. Don't run repair or converter apps against the originals: the fault is the card's media, and the fix is managed reading first, container repair second, on a copy. Note the pattern per file (freeze times, which clips refuse entirely) — it maps the decline — and expect verification by full playback, because for video nothing else counts.
The braid can be re-woven — email or call Belfast Data Recovery on 028 9002 0144 before another play-through.
Request a quote online →
Our case files are drawn from genuine enquiries received by our laboratory over the past ten years, anonymised to protect client confidentiality. Each one describes the diagnostic and recovery procedure our engineers apply to that fault, using the equipment listed.