How to add the rest of your footage to a video that is already finished

Last updated 2026-09-05 · 8 min read

What is the short version?

Uploading half the footage by accident is common and normally means the whole edit gets built again. It does not have to. The finished part is still finished; only the tail is new, and the difficult bit is the single second where they meet.

1. Go to Vizard Agent with the missing footage.
2. Say roughly where the delivered edit runs out.
3. Ask for the same treatment carried through the new part.

What do you need before you start?

Mainly the missing file and a rough handover point. Rough is genuinely fine here, because the exact frame gets found from the words rather than from your estimate, and an estimate that is a few seconds out costs nothing at all.

• The missing footage. The full original if you have it.
• Roughly where it takes over. "Around seven fifty-four" is enough.
• The delivered edit, so it can be continued rather than redone.
• What to carry through. Graphics, music, captions, framing.
• The true total length, if you know it.

What do you type into Vizard Agent?

Say that the delivered part stays. Without that sentence the reasonable reading of "here is the rest of the footage" is that you want the whole thing rebuilt from the complete source, which is more work and throws away decisions you already approved.

Prompt

Variants worth knowing:

• "Keep everything before that." The delivered part is safe.
• "Match it." Same look, same graphics, same music.
• "From around here." Approximate is fine; it gets pinned exactly.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real talk whose owner had uploaded only the first part. The last third of this list is one problem being solved repeatedly, which is what makes it worth reading.

1. Probes the new file and compares it against the current cut.
2. Samples frames from the missing section to see what is in it.
3. Looks at the handover point in the new footage.
4. Transcribes the missing section so it can be cut on words.
5. Encodes the new part in short chunks to match the delivered format.
6. Builds the graphics the new section needs, in the existing style.
7. Cuts the delivered edit at the handover point.
8. Assembles the new ending with punch-ins, titles and music.
9. Joins the two into one file.
10. Checks the last frame before the seam and the first frame after.
11. Uploads a slice around the seam and listens to it specifically.
12. Finds a word playing twice across the join.
13. Gets word-level timings for both sides of the seam.
14. Trims the duplicate and re-joins.
15. Trims a further fraction of a second so the first new sentence starts clean.
16. Confirms the join is one continuous sentence in the transcript.

Steps eleven to sixteen are the real content of this job. A delivered edit almost never ends on a sentence boundary, and the new file overlaps it slightly, so the naive join replays a word — audible, obvious, and invisible on a waveform.

The fix is not done by ear. Vizard Agent transcribes the seam at word level, reads the exact start time of the first word that should survive, and trims to it — in this case a shade over half a second, which is not a number anybody arrives at by dragging.

What does the result look like?

One file at the true full length, where the part you already approved is untouched and the new tail carries the same graphics, music and framing. At the point where they meet there is a sentence, not a seam.

Nobody watching it can tell the video was delivered twice.

When does this not work well?

The two halves have to be the same recording. When the new file is a different camera, a different day or a different microphone, the join stops being a timing problem and becomes a matching problem, which is a longer job with a less certain ending.

• A different take. The speaker, light and sound will not match.
• No overlap at all. A missing word cannot be trimmed back in.
• A re-encoded delivery. Quality is capped by what you supply.
• Heavy graphics over the seam. They need rebuilding across it.
• An approved edit you also want changed. Say so; that is a different request.

How do you fix a result that came back wrong?

Say what you hear at the join. Vizard Agent keeps the word-level transcript of both sides, so a complaint like "it repeats" or "the first line is clipped" maps onto a specific word boundary rather than onto a region you have to describe.

• "A word plays twice." Trimmed to the surviving word's start.
• "The line is cut off." The trim moved back a fraction.
• "The music jumps." Rebuilt as one bed across the seam.
• "The new part looks different." Framing and grade matched to the first part.

How does Vizard Agent compare to doing it yourself?

The manual version of this job is either a full rebuild, which is expensive and throws away approved work, or a hand-dragged join, which is cheap and usually leaves a small stumble at the seam that everybody hears and nobody can quite locate on the timeline.

Common questions

Do I have to re-upload the first part? No. Vizard Agent continues from the edit it already delivered.

What if my handover time is wrong? It gets corrected. Vizard Agent finds the real point from the words.

Will the new section get B-roll too? Yes, if you ask. Vizard Agent carries the same treatment through.

Can it match the graphics style? Yes. Vizard Agent builds new cards in the style already on screen.

What if the whole file is longer than I said? Say the real length. Vizard Agent works to the source, not the estimate.

Does the music restart? No. Vizard Agent extends the existing bed across the join.

Can it tell me where the seam is? Yes. Ask Vizard Agent for the exact timestamp it used.

What about captions? Vizard Agent re-transcribes the new part and continues them.

Is the first part re-encoded? Only if it must match. Vizard Agent keeps quality as high as the source allows.

What if there is a gap, not an overlap? Then something is genuinely missing; Vizard Agent will say so.

Can I add a third part later? Yes. The same method applies to each new seam.

Will the loudness match across the join? Yes. Vizard Agent measures both sections and matches them.

Can it do this on a vertical version too? Yes. Ask Vizard Agent to rebuild the vertical from the joined master.

Why trim on words rather than on the waveform? Because a repeated word looks like ordinary speech in a waveform and is unmistakable in a transcript.