Crisp for genealogists and family historians
Crisp is the restoration pass on family footage you have already digitised, run entirely on your own Mac. It straightens the combing on tape and telecined film before anything enhances the frames, lifts the single-frame dust off a cine transfer, evens out the brightness flicker, and keeps a black-and-white source black and white. It does not capture from tape, catalogue anything, transcribe the commentary you record, or remove long scratches — and it never uploads a frame of it.
Family history work has a deadline nobody sets. The tapes are decaying, the people who can identify who is in them are getting older, and there is exactly one copy of most of it.
That changes the priorities. Restoration quality matters less than getting everything captured and identified while it is still possible.
Capture first, restore later
This is the single most useful thing to know. Magnetic tape loses signal whether or not anyone plays it, and every playback wears it slightly. Software gets better every year; a tape from 1987 does not.
So capture the whole collection at the best quality you can manage now, even if you have no time to restore any of it. A digital file sitting on two drives is safe. A tape in a loft is on a clock. Restoration can happen in ten years; capture frequently cannot.
If a tape squeals or sheds powder, stop and get advice before playing it again. That is binder breakdown, and forcing it through a deck can destroy both the tape and the heads.
Identify while people still can
Sit down with the oldest relatives and record them watching the footage. Their commentary is frequently more valuable than the footage itself, and it is the part that genuinely cannot be recovered later. Names, places, dates, the year a house looked like that.
Do this before restoration, not after. It needs no processing and it is the piece with a hard deadline attached.
Then the restoration, in order
Deinterlace, denoise, upscale, colour. That sequence matters more than any individual setting. Nearly all tape and broadcast material is interlaced, and an upscaler reads the comb pattern as fine detail and sharpens it into something much harder to remove. Let the upscale handle the noise for the same reason: magnified noise is a worse problem than the original.
Be conservative with denoising on old film. Grain is part of how film looks, and stripping it out gives faces the waxy appearance that makes an over-restored archive obvious. You can always run a second pass; you cannot put the texture back.
Crisp does the deinterlacing step itself, as a pre-pass before anything enhances the frames, and it reads the answer off the pictures rather than off the file's own flag. That flag stops being true the moment a tape rip has been cut and re-encoded, which is the state most transfers reach you in: measured on one newsreel, the original MPEG-2 honestly reported interlaced fields, and the re-encoded copy claimed to be progressive while every frame was still combed.
It also keeps the two cases apart, which matters because they are not interchangeable. Genuinely interlaced video — a camcorder, a tape transfer — has two different moments woven into each frame, so a whole frame is rebuilt from each field pair and nothing is discarded. Film that went through a telecine has 24 real frames spread across 30 by repeating fields, so the duplicates are removed and the original frames put back: measured on a 3:2 fixture, 120 frames became exactly 96 at SSIM 0.9969 against the true original. Applying the wrong one to the other throws away a real frame in five, so Crisp checks the telecine answer on a sample before the render commits to it.
A shelf rather than a single clip is a batch job: point it at a folder and it works through videos and photos unattended, skipping anything it has already done, so an evening's queue can run overnight. Batch is one of the paid capabilities rather than part of the free tier.
Cine film: dust, flicker and black-and-white
Reels of 8mm and 16mm arrive with a different set of faults from tape, and three of them are handled by ordinary filter passes rather than by a model — no weights, no extra render time, and they work alongside every upscale preset.
Single-frame dust is the one worth knowing about. A speck of dirt sitting on one frame and not on its neighbours can be voted away by comparing each pixel across a short window of frames. Measured on dirt planted over real film frames, that removes 95 to 100 per cent of it — and a blotch held in the same place for three frames correctly survives. That second result is the one that makes the first trustworthy: a filter that merely blurred would have "cleaned" the sustained dirt too. Ask for it in plain English — "remove the dust", "get the specks off this" — and it goes on the plan before you confirm it.
Brightness flicker, the exposure pulse that makes old film breathe, is evened against a moving average: measured at an 87.6 per cent reduction in frame-to-frame brightness jitter, while a deliberate fade-in came through intact. Keeping the fades somebody meant is the part that matters, because flattening them would ruin footage that was never broken.
And a source with no colour does not acquire any. Crisp checks for that itself, sampling three windows spread through the clip rather than only the opening, so a black-and-white title sequence in front of colour footage is not mistaken for a monochrome reel — any colour found anywhere vetoes the whole clip.
Scratches are the exception, and Crisp says so instead of trying. A scratch persists across many frames, so the cross-frame comparison has nothing clean to vote with and smears the picture rather than healing it. Ask for scratch removal and the app declines and offers the dust pass instead, which is a better answer than a plausible-looking export you would have to spot the damage in later.
What cannot be recovered
Worth knowing before you spend a weekend on it. Tracking tears, dropouts where the signal was never read, and timebase instability all have to be fixed at capture with a properly aligned deck, not afterwards in software. A shot that missed focus has no detail to recover, and an upscaler will invent plausible features rather than restore real ones, which on a face you are trying to identify is worse than leaving it soft.
Blown highlights are gone. Audio that was never recorded is gone. Accepting that early saves the time better spent on capture.
Where Crisp is the wrong tool
It does not capture from tape, so you still need a deck, a capture device and ideally a timebase corrector. It does not catalogue, tag or manage metadata, which for a serious archive matters as much as the footage. And it does not transcribe speech, so the commentary you record still needs transcribing elsewhere.
What it does is the restoration pass, on your own machine, on footage of your family that never goes to anyone else's server. For this particular use, that tends to be the point rather than a feature.
Family archives, in detail
Should I restore before or after digitising everything?
Digitise everything first, without exception. The tape is degrading on a clock and software only improves. A complete set of raw captures is far more valuable than a handful of beautifully restored clips and a box you never got to.
How much should I denoise old film?
Less than feels right. Grain is part of the image, and heavy denoising produces the waxy, plastic faces that give an over-processed archive away. Remove enough that it reads cleanly and stop there.
Can Crisp fix tracking lines and dropouts?
No. Those are places where the signal was never read, so there is nothing in the file to work from. They need a properly aligned deck and ideally a timebase corrector at capture time.
Is upscaling a face reliable for identification?
Be careful here. An upscaler predicts plausible detail rather than recovering real detail, so a sharpened face is partly invented. For identification purposes the soft original is more honest evidence than the crisp reconstruction.
Can Crisp clean the dust and scratches off 8mm film?
The dust, yes. Specks and dirt that appear on a single frame are removed by comparing each frame against its neighbours, measured at 95 to 100 per cent of planted dirt, while a blotch held for three frames correctly survives. Scratches, no. A scratch persists across many frames, so there is nothing clean to compare it against, and Crisp declines the request rather than smearing the picture.
Related
See also: restoring old home videos on a Mac, which covers the general tape workflow this page assumes.
Download Crisp for Mac Free to try, one-time $129 to remove the watermark. Runs entirely on your Mac.