HANDS-ON RESULTS / SIX TOOLS

Provisional · single tester

Hands-on test results: six AI clippers

Six AI clipping tools, one results table, and a plain note on which scores can be compared.

By Christopher Brown · Last updated

What a Scrutator page covers and leaves out is set out once on our methodology page.

All scores here are provisional: one tester, one round per tool. Four tools completed the full protocol (three sources, two runs each) and can be compared directly. Descript is a partial sample and Captions could not be graded. The method is on How we test.

Most lists of AI clipping tools are written by companies that sell one of them. Scrutator does not sell a clipping tool and does not use affiliate links. This page collects every hands-on test we have run so far in one place. Every number below is copied from the tool’s own research file on this site, or from the run log you can download at the bottom. Where we do not have a number, the table says so rather than estimating one.

Hands-on test results for all six AI clipping tools

Read the “Comparable” column before the score column. A score only means something next to another score if both came from the same sources, the same number of runs and the same grading rules.

Hands-on results for six AI clipping tools, October 2026 (provisional, single tester)
ToolScore / 100TestedSources × runsClipsAccepted / repairable / rejectedEst. repair timeCost per clip (our calculation)Comparable
KlapPricing →82.43 sources × 2 runs4121 / 20 / 074 min$0.14 per usable, $0.27 per accepted (Basic, billed yearly)Yes
PictoryPricing →79.03 sources × 2 runs5437 / 17 / 019 min$0.54 per usable, $0.78 per accepted (Starter, monthly)Yes
Opus ClipPricing →73.53 sources × 2 runs7727 / 47 / 3205 min$0.049 per usable, $0.133 per accepted (Starter)Yes
VugolaPricing →51.63 sources × 2 runs150 / 15 / 087 min$1.12 per usable; no clip accepted as delivered (Starter)Yes
DescriptPricing →76.82 sources × 1 run (partial)10512 min$0.36 per usable, $0.72 per accepted (credit share of Hobbyist)No — partial sample
CaptionsPricing →No score3 different sources × 2 runs28Not gradedNot measuredNot calculatedNo — not graded

Accepted = publishable as delivered. Repairable = usable after a fix, with the fix time estimated per clip. Rejected = not usable. Cost per clip is our own calculation from the plan named in each row; plans and billing periods differ, so read it as a rough guide, not a shared unit.

Which scores you can compare

Directly comparable: Klap (82.4), Pictory (79.0), Opus Clip (73.5) and Vugola (51.6). All four ran the same three 6:58 rights-cleared sources — a single-speaker tutorial, a two-speaker interview and a screen-sharing demonstration — twice each, six jobs per tool. Every clip was graded against the same definition of a usable clip, written before any results were seen. Opus Clip and Vugola ran on October 19, 2026; Klap and Pictory on October 29, 2026.

Not directly comparable: Descript (76.8). Descript’s Create Clips feature costs 30 AI credits per run, and the account ran out after two runs. We completed the tutorial and the interview once each, with no screen-recording source and no second run. 76.8 is therefore a partial-sample score. It sits between Pictory and Opus Clip on the number line, but it was earned on two jobs, not six, so treat the ranking as a hint rather than a result.

Not graded: Captions. Captions’ Clips mode ran six times on October 30, 2026 and returned 28 vertical clips. Downloading a clip required a paid subscription, and the in-browser preview did not play on our test machine, so we could not check caption errors, framing, resolution or watermark. Captions also ran on three different sources, all short talking-head videos (3:41 to 4:07), after earlier uploads of our standard sources timed out. There is no Captions score, and we will not estimate one.

Each tool’s page and the research directory carry the same test-status label taken from this table: “Full protocol (comparable)”, “Partial sample” or “Ran, not graded”. Pages with no run of their own are labeled “Documentation checked, not hands-on”.

What each AI clipping tool did

Klap — 82.4, highest on the full protocol

Klap returned 41 clips: 21 accepted, 20 repairable, none rejected. It was strongest on the single-speaker tutorial, where 18 of 20 clips were ready to post. The interview and the screen demo were harder: most clips there needed trims or framing fixes, which is where most of the 74 estimated repair minutes went. Exports were 1080×1920 vertical with no watermark on the Basic plan. Details: Klap pricing and test results.

Pictory — 79.0, fewest repair minutes

Pictory returned 54 clips: 37 accepted, 17 repairable, none rejected, with only 19 estimated repair minutes in total. Two things to know: it produced 16:9 landscape clips (we saw no 9:16 option in this flow), which avoids vertical-crop mistakes, and exports on the trial carried a “Created with Pictory” badge and outro card. The listed clip length was usually several seconds shorter than the actual file. Details: Pictory pricing and test results.

Opus Clip — 73.5, most clips, most repairs

Opus Clip returned the most clips — 77 — of which 27 were publishable as delivered, 47 usable after a fix and 3 rejected. Its repair load was the largest in the test at about 205 minutes, many of them trims at the start or end of a clip (14 of the 24 repairable talking-head clips). On cost per usable clip it was the cheapest of the four comparable tools. Details: Opus Clip review and Opus Clip pricing.

Vugola — 51.6, best captions, weakest value

Vugola returned 15 clips, none publishable as delivered and all 15 usable after a fix, with about 87 minutes of repairs. Its burned-in captions were the most accurate in the October 19 round, but the cost per usable clip, about $1.12 on Starter, pulled its value score down. Details: Vugola AI review, Vugola AI pricing and Vugola vs Opus Clip.

Descript — 76.8, partial sample

Descript’s Create Clips returned 10 vertical 720×1280 clips across two jobs: 5 accepted, 5 repairable, about 12 minutes of repairs. Exports on the plan we used carried a watermark, one export failed with a font error before succeeding per clip, and framing was inconsistent. Its transcript-based editor makes caption fixes easier than in pure clipping tools. Details: Descript pricing, Descript alternatives and Descript vs Riverside.

Captions — clip counts only

Captions was consistent in one respect: the same source gave the same number of clips on both runs (3, 5 and 6), although titles and its own “viral scores” changed between runs. An earlier October 29 attempt failed with “No speaker detected” and upload timeouts. Details: Captions AI pricing.

How to use these results

Whatever you pick, run your own footage before paying for a year. Our sources are three videos; yours may behave differently. For tools we have only read about, see best AI clip makers, Opus Clip alternatives and the AI video pricing tracker.

Limits of this test

One person graded every clip, so accepted versus repairable is a judgment call made by one person, applied with the same written rules. Each tool ran in one round; vendors update models often, and a second round could move any score. Cost figures depend on the plan and billing period named in the table. Repair minutes are estimates, not stopwatch times for every clip. We will add rounds and update this page with a dated note rather than silently replacing numbers.

Download the run log

The per-run log for the October 29–30 tests (Klap, Pictory, Descript and Captions) is available as a CSV: one row per job, with settings, clip counts, accepted / repairable / rejected, repair minutes, caption error counts and export format. Skipped Descript runs and ungraded Captions runs are included and labeled. Opus Clip and Vugola per-source results are in the tables on their own review pages.

Download run log (CSV) ↓

Questions about these hands-on test results

Which AI clipping tool scored highest?

Among the four tools that completed the full protocol, Klap scored highest at 82.4, followed by Pictory at 79.0, Opus Clip at 73.5 and Vugola at 51.6. Descript’s 76.8 comes from a partial sample and Captions has no score, so neither belongs in that ranking.

Why is Descript’s score not ranked with the others?

It was earned on two jobs instead of six, with no screen-recording source and no repeat run. Repeat runs are where variability shows up; without them a score can look steadier than it is. We show the number so you can see where it landed, and label it so you do not over-read it.

Why does Captions have no score?

Our score needs every dimension graded: clip quality, captions, automation, value and usability. Downloading clips required a paid subscription and the preview would not play on our test machine, so we could grade none of the clip-level dimensions. A missing dimension means no overall score, not a zero and not an estimate.

Did the vendors pay for these tests?

No. Vendors do not pay for coverage, do not see results before publication and do not set scores. There are no affiliate links on this site.

How often will this page change?

Whenever we finish a new round or a new tool. Each change is dated, and earlier figures stay on the tool’s own page with the date they were measured.

About the author. Written by Christopher Brown; published by Scrutator Team, Hong Kong, which is responsible for this page, its method and its corrections. Questions or corrections: contact us.

Read the full method ↗