Cutroom vs Descript: one edits the recording, one hands back a finished vertical ad

Their own line is that video editing is as easy as typing. A transcript arrives with the file. You edit the video by editing the text, and on the enterprise tier they describe it as editing video by editing a doc.

On a long recording that is genuinely the better interface. An hour of talking is faster to read than to scrub. Deleting the paragraph where the guest lost the thread takes a second.

It is not only a transcript either. There is a real timeline underneath with precision controls. When the machine gets a cut wrong, you fix it yourself instead of asking for another attempt.

It also records. Screen recording, remote recording rooms, multitrack audio. It translates and dubs into 30 or more languages with a proofreading step. A Create Clips tool picks moments out of a long recording for short social cuts.

The sources it takes are large. Upload limits vary by plan under a published hard ceiling of 50 GB a file, across MP4, MOV, WEBM, MKV and M4V containers. Output runs 720p to 4K, at 1:1, 9:16, 16:9 or a custom ratio, at frame rates from 23.98 to 60.

That is a serious product and this page is not going to pretend otherwise. It is also a different weight class from what follows.

Editing text removes what you said twice. Rearranging what is left into ad shape is work you do afterwards, on the timeline underneath, and Descript gives you a real one to do it with.

So the text pass hands back your recording, minus the parts you deleted. Everything past that is a manual job, which on an hour of tape is exactly the right trade to offer.

It is fair to say what Descript does publish here, because it is more than nothing. The Create Clips tool named above uses their own AI to pick the moments most likely to go viral, and that is a genuine selection decision aimed at long sources.

Quick Design formats and adds b-roll with a click, and Descript says so plainly. It publishes where those clips can come from too: generated from a text prompt, pulled from its royalty-free stock library, or your own uploaded material. What it does not publish is which phrases the automatic pass covers, or any rule holding your face on screen for a set share of the runtime, so coverage stays something you check by watching.

The free plan has a shape worth knowing before you test on it. Exports from a free drive carry a Descript watermark across the middle, and free export is capped at 720p. Upgrading clears it, but anything already exported has to be exported again.

Media minutes are the other shape. They are spent when you upload or record, not when you transcribe, so the free 60 are a budget for what goes in. Still images cost a second each. The 100 AI credits on the free plan are a one-time allocation rather than a monthly one.

Upload one talking-head take of up to three minutes. It gets transcribed, and a full cut gets drafted before you decide anything.

Slack out. Pace set. B-roll already sitting on the phrases that need showing. A headline drafted on the opening. Captions styled and burned in. Music optional, sitting under the voice at a level you set.

Then you direct it by reading. Highlight a phrase and a clip lands over exactly those words. Delete a line and the cut rebuilds around the gap. Change the pace and the whole piece re-cuts.

There is no timeline and no layer stack. A talking-head ad is decided by words rather than pixels, so the interface is the transcript and the compiler works out every boundary in time.

Eleven moves is the whole vocabulary. Emphasise the word the claim turns on. Trim the opening. Swap a clip, or search millions of free ones. Fix a word the transcription misheard before it reaches an export.

One project produces one finished vertical. Coverage is capped by pace, 40 percent on chill, 45 on normal and 52 on fast, so your face holds the frame for most of the ad whatever the footage does around it.

One. B-roll lands on the words you name. Highlight a phrase and a clip covers exactly those words, so coverage becomes a reading decision instead of a hunt for the right second on a playhead.

Two. Your own footage gets first pick. Your product, your premises and your hands sit in the workspace library and drop onto the words that describe them. Descript takes your own material as b-roll too, and its own documentation shows an uploaded slide deck used that way. The difference is who places it, and on which sentence.

Three. Presenters from a photo are a draw, and pretending otherwise would cost this page its credibility. Descript documents animated avatars built from an uploaded photo or a text prompt, lip-synced to a voice, running on Kling models and spending AI credits. It lists photo avatars on the Business row of its pricing page.

Here the same idea is priced by the minute. A photo you supply plus a voice-over becomes a lip-synced presenter at 350 credits a minute of render, and that presenter drops into the same pipeline that cuts, covers and captions the ad.

Four. The file you export while trying it is the file you would have paid for. Descript's help pages say a free drive stamps its watermark in the bottom-right, caps export at 720p, and needs anything already exported to be exported again after an upgrade. Three finished videos are free in your first seven days, with no card.

Five. The second version is cheap. Re-cutting a take you already batched costs only the export at 20 credits per output minute, so trying a different opening is an afternoon rather than a budget conversation.

Their pricing is theirs to state, so read it at the source. What they publish is five plans, Free through Enterprise, with Enterprise priced on request and annual billing advertised as saving up to 35 percent.

Ours in full, because it is ours to state. A batch covers transcription, the director pass and the b-roll search at 100 credits. Export is 20 credits per output minute, so a one-minute ad is about 120.

Basic is $39.99 a month for 2,500 credits, which is roughly twenty one-minute ads. Premium is $79.99 for 5,000. Top-ups are $39.99 for 2,500 when a month runs hot. The first three videos are free, over 7 days, with no card.

Pick Descript if the source is long, if you need a timeline to fix a cut by hand, if you record remote guests, or if you ship square and horizontal as well as vertical. Its enterprise tier also carries the answers a procurement process asks for, including SOC 2 Type II, SSO, SCIM and audit logs.

Pick Cutroom if the job is one filmed read becoming one vertical ad, this week and next week, at a volume where the hour per edit is the actual bill.

The facts a buyer needs, flatly. Three minutes is the upload ceiling. One speaker. Vertical MP4 out and nothing else. No recording of any kind here, no dubbing, no timeline underneath, and no frame-level nudging of anything.