How to get photographs out of a video you already have
Upload the video and name the moments you want stills of. Vizard Agent works through the whole recording as numbered contact sheets, extracts detailed frame series around each moment, compares the candidates side by side at full resolution, and finishes the ones that win.
What is the short version?
Somebody filmed the day and nobody photographed it. A video at thirty frames a second contains thousands of potential photographs, and the only reason you have none is that finding the good ones means looking at all of them.
- Go to Vizard Agent and upload the video.
- Name the moments you want photographs of.
- Say how many you want, and what they are for.
What do you need before you start?
The video at the best quality you have, and a list of moments. Vizard Agent compares candidates at full resolution, so the difference between a phone export and the original file is the difference between a usable photograph and a soft one.
- The original file. Not a compressed re-upload.
- The moments. The ring, the toast, the finish line.
- The count. Ten strong ones, or five.
- The use. Print, a frame, a social post.
- The people. Whose faces have to be right.
What do you type into Vizard Agent?
Say what the photographs are of. Vizard Agent finds the moments itself, so naming them — rather than describing the feeling you want — is what gets you stills of the things you will actually want to keep.
Prompt
Variants worth knowing:
- Name each moment. It builds a candidate set per moment.
- Say the count. It stops at strong ones rather than filling a quota.
- Say what they are for. Print needs the sharpest frames.
What does Vizard Agent actually do?
Here is the order Vizard Agent worked in on a real proposal video filmed while paragliding. Nearly half of the session is spent simply looking at frames, because choosing a photograph is a judgement about faces and timing, and there is no measurement anywhere that substitutes for having looked.
- Analyses the video and checks its technical specifications.
- Extracts preview frames to see the scenes it contains.
- Reviews the frame grid for an overview of the whole thing.
- Extracts frames across the entire duration and builds numbered contact sheets.
- Reviews the contact sheets in three passes — frames 1 to 20, 21 to 40, 41 to 59.
- Extracts detailed frame series around each key moment.
- Builds comparison contact sheets for each scene separately.
- Reviews the landscape frames, then the flight, then the embrace.
- Reviews the proposal frames, the bouquet, the ring going on, and the kiss.
- Copies the candidate frames at full resolution.
- Builds pairwise comparisons of the shortlisted candidates.
- Compares each pair directly to choose between near-identical frames.
Step eleven is the part that produces a photograph rather than a screenshot. Two frames a tenth of a second apart look identical in a contact sheet and completely different at full size — one has the eyes open and the other does not.
What does the result look like?
What comes back is a set of finished photographs rather than a folder full of exported frames: each one chosen from a shortlist built for that particular moment, compared directly against its immediate neighbours at full resolution, and then processed and graded.
The value is in what was rejected. For every photograph delivered there were four near-identical candidates a few frames either side, and the difference between them is exactly the difference between a picture somebody frames and one they scroll past.
When does this not work well?
A single video frame is not the same thing as a photograph taken by a camera that was exposing for a still image, and there are hard physical limits on what can be recovered from one. These are the cases where the footage simply will not give you a picture.
- Motion blur. Video exposes for motion, not for stillness.
- Low resolution. A 1080p frame is a small photograph.
- Compressed uploads. The artefacts show at full size.
- Fast action. Every frame may be blurred.
- Print sizes. Ask before promising anyone a large print.
How do you fix a result that came back wrong?
Say which moment is wrong and what is wrong with the frame you were given. Vizard Agent keeps every candidate set and all the comparison sheets it built, so a different choice comes straight from the existing shortlist rather than another pass through the whole video.
- "His eyes are closed." Swapped for the neighbouring frame.
- "Too blurred." Replaced from a steadier moment nearby.
- "I want one of the landscape too." Added from its own candidate set.
How does Vizard Agent compare to doing it yourself?
By hand this is scrubbing frame by frame and exporting stills until you lose patience, usually around the third moment. Vizard Agent builds candidate sets for every moment and compares them at full resolution rather than in a timeline preview.
| By hand | Vizard Agent | |
|---|---|---|
| Coverage | The moments you remember | Whole video in numbered sheets |
| Choosing | From a small preview | Full-resolution pairwise comparison |
| Near-identical frames | Picked at random | Compared directly against each other |
| Finishing | Exported raw | Processed and ready to use |
Common questions
How many photographs can it get? As many as the video supports. Vizard Agent stops at strong ones rather than filling a number.
What resolution will they be? The video's own. Vizard Agent cannot add detail that was never recorded.
Can they be printed? Small prints, usually. Vizard Agent will tell you when a frame will not hold at size.
Does it fix motion blur? Not really. Vizard Agent chooses the steadiest frames rather than sharpening blurred ones.
Can it process them? Yes. Vizard Agent grades and finishes the chosen frames rather than exporting them raw.
What if I only want one moment? Say so. Vizard Agent builds a bigger candidate set for that one moment instead.
Can it find moments I did not name? Yes. Vizard Agent surveys the whole video and can suggest ones you missed.
Does it work on old footage? Yes, within its resolution. Vizard Agent handles home video the same way.
Can it crop them? Yes, to any shape. Vizard Agent re-frames each still for where it is going.
What about a group photo nobody took? If everyone is in one frame, it can be found. Vizard Agent looks for the moment they all faced the camera.
Why does full-resolution comparison matter so much? Because the differences that decide a photograph — a blink, a half-turned head, a hand in the way — are invisible at thumbnail size and obvious at full size.
Can I get contact sheets to choose myself? Yes. Vizard Agent can hand over the shortlists rather than deciding for you.
Does it check its choices? Yes. Vizard Agent compares each finalist directly against its neighbouring frames at full resolution before finishing it.
Can it get stills for a thumbnail? Yes. Vizard Agent picks frames with room for text and a clear subject when you say that is the purpose.