If you are picking an AI clipping tool, the comparison everyone publishes is a feature grid: captions, auto-reframe, b-roll, scheduling, export resolution. Nine rows, eight columns, ticks almost everywhere.
That grid cannot express the thing that actually decides what you pay, because these products meter work on units that do not convert into each other.
Here are the five I found across eight tools:
| Meter | What it counts | Cost scales with |
|---|---|---|
| Flat per job | one clipping run, any length | number of runs |
| Per source minute | length of the video you upload | input duration |
| Per output clip | clips the tool returns | output count |
| Media hours | hours of media processed | input duration |
| Recording hours | hours you record on the platform | capture volume |
A per-source-minute tool and a per-output-clip tool cannot be compared on price at all until you fix both the length of your input and the number of clips you expect out. Change either and the ranking inverts.
The measured run
Four of the eight I ran the same file through: one 35-minute conference panel, same source, same settings where settings existed.
| Tool | Clips found | Shown without paying | Clip length | Cost of that one job |
|---|---|---|---|---|
| Submagic | 31 | 31 | short-form | $1.53 |
| OpusClip | 32 | 32 | short-form | $2.03 |
| Vizard | 30 | 10 | 0:32-0:56 (the ten) | $0.85 |
| CapCut | 8 | 8 | 1:26-3:58 | $0 |
Costs are each tool's entry paid annual rate applied to that single job, except CapCut, which charges nothing for it. Submagic's higher tier works out cheaper per job than its entry one, at about $1.09, which is its own small lesson about reading a ladder rather than a sticker.
Three things in that table are invisible on a feature grid.
The flat-fee tool is the expensive one here and the cheap one later. Paying a fixed price per run is bad value on a 35-minute source and good value on a three-hour one. The crossover sits around an hour of source video. If your median input is a 20-minute interview, a flat fee is a tax; if it is a full podcast episode, it is a discount.
One tool's free tier withholds most of what it found. It reported thirty clips and delivered ten, with the rest behind an unlock banner, and the credits had already paid for analysing all thirty. Defensible, but it means "clips found" and "clips you can use" are different columns, and only one of them is on the marketing page.
One tool's "clips" are not short-form. Eight outputs running 1:26 to 3:58, totalling about three quarters of the source, is chaptering. That is a genuinely useful thing to have, and it is free, but if you planned to post them you have received eight editing jobs rather than eight assets.
What I would actually do
Work out your input shape first, then pick the meter, then look at features.
- Long sources, few of them: flat per job wins past roughly an hour.
- Long sources, many of them: per source minute, and compare the per-minute rate rather than the monthly sticker. On our run the same unit varied by about 2.4x between two tools.
- Short sources, many clips out: avoid per-output-clip pricing unless you bin most of what you get; it is the only meter that charges you for the tool's own generosity.
- You are already editing anyway: media-hours tools fold clipping into a bill you are paying regardless, which is cheaper than it looks on a comparison table.
- You record on the platform too: recording-hours pricing can make the clipping effectively free, and can also strand you if you ever record elsewhere.
Limits on this, stated plainly
One source file, one run per tool. It is a measurement, not a benchmark, and a different video would move the numbers.
Four of the eight tools I actually ran. The other four - Klap, Descript, Captions and Riverside - I priced and specced from their live pages and did not test. Where I describe what those produce, I am reporting a vendor claim, not a finding.
And one price caveat worth repeating: one of these vendors' annual rates read differently on different days, a lower figure on one afternoon and a higher one on three other reads across two days, so the derived per-job cost above uses the higher figure as the rule and treats the lower as promotional.
The full comparison, with the per-tool write-ups, the free-tier table and the break-even maths, is in our roundup of Submagic alternatives.
Top comments (0)