How to hit a file size limit without shortening the video

Last updated 2026-09-10 · 8 min read

What is the short version?

A file size limit and a runtime together decide your bitrate — there is no third option. Once you know that number the question stops being "how do I make it smaller" and becomes "where do I spend what I have".

1. Go to Vizard Agent with the cut.
2. Give the limit in megabytes and the length that must stay.
3. Say what has to survive — text, faces, fast motion.

What do you need before you start?

The limit and the length, and a view on what matters. That last part is what turns a mechanical encode into a decision: text and faces need detail, a dark slow shot does not, and knowing which you have lets the budget go where it shows.

• The size limit. In megabytes.
• The runtime. And that it must not change.
• What must stay sharp. Captions, faces, a logo.
• The format. Resolution affects everything here.
• Whether audio quality matters. It costs some of the budget.

What do you type into Vizard Agent?

Give both numbers together. A size limit on its own invites shortening the video, and a runtime on its own invites ignoring the limit — stating both makes it clear which variable is left to move, which is the picture quality.

Prompt

Variants worth knowing:

• "Under 100 MB but over a minute." Both constraints at once.
• "The captions have to stay sharp." Where to spend the budget.
• "Tell me what it came out at." Verification, not assurance.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real edit that had to come in under a hard size limit without losing any of its length. Most of the list is arithmetic that happens before a single frame is encoded.

1. Takes the limit and the runtime and derives the available bitrate.
2. Checks whether that bitrate is workable at the requested resolution.
3. Proposes a resolution if the numbers do not fit.
4. Looks at the footage to see where detail actually matters.
5. Allocates more of the budget to the detailed sections.
6. Sets the audio at a sensible share rather than the default.
7. Encodes the file to the derived target.
8. Measures the finished file against the limit.
9. Adjusts and re-encodes if it came out over.
10. Checks the captions and faces at full size in the result.
11. Confirms the runtime is unchanged.

Step one is the whole method. Size divided by length is a bitrate, and every other decision — resolution, how much goes to audio, whether the fast sections can afford their detail — follows from that number rather than being guessed at.

Step four is what stops this being a purely mechanical encode. Looking at the footage first shows where the detail is doing work — captions, faces, a logo — and where a shot is dark and slow enough that nobody will ever notice it was given less, which is exactly where the savings should come from.

Step eight is not a formality. An encoder given a target bitrate does not always deliver a file under the limit, particularly on material with a lot of movement, so the finished file gets measured and re-encoded if it overshoots.

What does the result look like?

A file under the limit, at the length you needed, with the parts that carry meaning — the captions, the faces — still sharp, and any softness pushed into the shots where nobody was going to look closely anyway.

The number is confirmed rather than hoped for.

When does this not work well?

Sometimes the numbers simply do not allow it. A long video at a small limit leaves a bitrate that no allocation Vizard Agent makes can rescue, and at that point the honest options are a lower resolution, a shorter cut, or a bigger limit.

• Very long at a very small limit. The arithmetic refuses.
• High-motion footage. It costs far more bitrate.
• Fine detail throughout. Nothing to sacrifice.
• Lossless requirements. Not compatible with a size cap.
• Limits set by an old system. Ask whether they still apply.

How do you fix a result that came back wrong?

Say whether the problem is the size it came out at or how it looks. Vizard Agent keeps the bitrate calculation and the allocation it made across the video, so a fix moves budget between sections or drops the resolution a step rather than starting the encode over from scratch.

• "It came out over." Re-encoded and re-measured.
• "The captions look soft." More budget to those sections.
• "It is blocky in the fast bits." Motion given a larger share.
• "Can it be smaller still?" The trade shown before it is made.

How does Vizard Agent compare to doing it yourself?

By hand the usual approach is a quality slider and a few attempts, which converges eventually and spends the budget evenly across the video. The file is also generally trusted to be under the limit because the encoder was told a number.

Common questions

Can it keep the length? Yes. Vizard Agent moves picture quality rather than runtime.

What if the numbers do not work? Vizard Agent says so and offers a lower resolution instead.

Will the captions stay readable? If you say so. Vizard Agent spends the budget there.

Does it lower the resolution automatically? Only if it has to, and Vizard Agent tells you when it does.

How much goes to the audio? A deliberate share. Tell Vizard Agent if the audio matters more.

Will it check the actual file? Yes. Vizard Agent measures it against the limit.

Can it hit a smaller limit? Yes, up to the point the arithmetic refuses.

What about a platform's own re-encode? That happens after you; a smaller file gives it more to work with.

Can I have two versions? Yes. Vizard Agent delivers one for the limit and one at full quality.

Does it affect the frame rate? Only if you allow it. Tell Vizard Agent whether that is on the table.

Will fast motion look worse? It costs the most, so Vizard Agent allocates it deliberately.

Can it tell me the bitrate it used? Yes. Ask Vizard Agent for the numbers.

Does this change the edit? No. Vizard Agent re-encodes rather than re-cuts.

Why does size divided by length matter so much? Because it is the entire budget, and every quality decision in the file is a way of spending it.