As of April 2026 there are more than 40 publicly known AI editors for short-form video, up from six in 2024. Something like nine in ten are wrappers around the same three parts: a Whisper-class transcription model, a GPT prompt that says "pick three viral moments", and a captioning template. What actually differs between them is the marketing copy and the pricing tiers.

That's why "I tried three AI editors, none of them clicked" is such a common line in coach and podcaster circles. Most tools are tuned for one kind of source, usually the 90-minute podcast into six reels, and fall apart on everything else. Here's a practical look at four that keep coming up in customer conversations: CapCut Auto-Cut, Opus Clip, Vizard, and our own Lotima. Honest about our own gaps, no affiliate links, no feature-checklist tables.

What an AI editor actually does in 2026

Before comparing them, it helps to be clear on what these tools actually automate:

1. Transcription turns the audio into text with timestamps. Since 2024 this has been a solved problem: Whisper-large-v3, Deepgram Nova-2, and ElevenLabs Speech-to-Text all hit 95 to 98% word accuracy on clean speech across more than 50 languages. The tools start to differ after this step, not at it.
2. Moment selection is choosing which fragments will work in short form. This is where the real difference lives. Chunks can be ranked by voice energy, by trigger-word matches, by narrative completeness (does the fragment have a setup, tension, and payoff on its own), or by hook strength (does the first sentence stand alone). Most tools do one of the first two, which is cheap and fast but gives you a "loudest moments" cut that often makes no sense on its own.
3. Reframing turns 16:9 into 9:16 so the speaker stays framed instead of getting a dumb center crop. All four handle this. Where they split is scenes with two or more people on camera.
4. Captioning burns in subtitles with word-level highlighting. The differences here are small; everyone ships more or less the same template stack.
5. Pacing cuts silences, evens out phrase length, and layers in B-roll. This is the hardest part. Trim pauses with taste and you get the AI edit without the AI feel; trim them clumsily and a thoughtful talk turns into rapid-fire babble.
6. Brand consistency applies your preset (font, color, watermark, intro and outro card) without you touching it.

The most common mistake is to compare tools by their feature checklist. If all four do captioning and reframing, those tell you nothing. The real differences sit in moment selection, pacing, and brand consistency, and that's where the hours you save each week actually come from.

CapCut Auto-Cut

CapCut is ByteDance's full video editor with an Auto-Cut feature built in. Its strength is everything around it: millions of ready-made templates, direct publishing to TikTok and Reels, a free mobile version that's hard to beat. Auto-Cut shipped in 2023 and has since split into Auto-Cut, Auto-Reframe, and Long Video to Shorts modes.

Where it works: a solo creator who already has a 10-minute clip and wants to ship one reel with a clean three-beat structure (three lines, cut, outro). Auto-Cut reliably gives you a 70-second draft in a minute or two, and you finish it off in the desktop timeline. If you already live in CapCut, there's no reason to leave.

Where it falls apart: long source material, the 60-minute-plus lectures and podcasts. In Long Video to Shorts mode, Auto-Cut reliably grabs the first three or four loud fragments with no sense of how the talk is built. You end up with three near-identical clips from the first 20 minutes while the real payoff sat between minutes 35 and 50. It also doesn't separate speakers, so if you're a podcaster with a guest, every caption pins everything on one voice.

Pricing: free on the basic tier with a watermark, and CapCut Pro at $7.99 a month for watermark-free export and unlimited Auto-Cut. Cheap, but it isn't built for batching.

Opus Clip

Opus Clip is built around one job: turning a long podcast or interview into vertical clips with auto-captions and a ClipAnything prompt. Since launching in 2023 it's pulled in roughly 10 million creators and consistently ranks at the top of long-to-short comparisons.

Where it works: a podcaster with one or two guests uploads an hour-long episode and gets eight to twelve clips ranked by a viral-potential score. Speaker labeling works well enough, and the brand templating is flexible. If your main format is dialogue video, Opus is the most predictable choice out there.

Where it falls apart: the solo talking-head, the everyday coaching reel. Opus is tuned to find dialogue beats, line then reaction then punchline, and on a monologue it has nothing to grab, so it falls back on stand-ins like volume and speech rate. You get cuts landing in odd places: a coach clipped mid-sentence because their voice happened to dip at that second. Lectures have the same problem.

Pricing: free up to 60 audio minutes a month, Starter at $19 a month for 300 minutes, Pro at $29 a month for 1,200. Probably the best long-to-short tool for a podcaster with one host and one guest, but not a generalist.

Vizard

Vizard is Opus's most direct competitor, aimed hard at agencies and B2B teams. What sets it apart is native integration with Zoom, Loom, and Google Meet: one click after a call ends and your meeting becomes marketing clips.

Where it works: a marketing team recording five-plus hours of webinars and expert interviews a week. Vizard really does shrink the work down to "after the call, clips ready in an hour". Speaker labels, auto-translation into more than 30 languages, brand templates, all solid.

Where it falls apart: its moment selection is noticeably weaker than Opus on long podcasts, and its pacing on a talking-head is weaker than CapCut. Vizard puts more into the integrations than into the selection itself, and you can see it in the clips, which often look like the middle third of a Zoom call with the ends trimmed, no real hook.

Pricing: 720 free minutes one time, Creator at $20 a month, Pro at $40, Enterprise on request. If most of your work is Zoom webinars, it pays back fast. For a solo coach filming on an iPhone, it's overkill.

Lotima

We're building Lotima on a different premise: the whole edit happens in a chat with an agent, and the agent has to hand you a clip that carries a complete thought, not just a loud moment. You upload a source in any of three formats (talking-head, lecture, podcast), ask for a reel, and the software transcribes the footage, reads the narrative, picks the strongest moment, then cuts, captions, and paces it. Before it renders, the agent shows you its edit plan and explains why it chose each cut. If you disagree, you say so in chat: change the caption style, fix a misheard word, tighten the pacing, swap the music.

Two things follow from that, and it's fair to call them limitations. First, one pass gives you one reel, not a ranked batch of twelve. You get the next reel by asking for another moment from the same source, a few minutes of chat each. Second, there's no manual timeline. If you want to drag clip boundaries by hand, that's on our roadmap, not in the product today.

Where it works: on a mixed diet of lecture, interview, and solo talking-head, where you care that each clip stands on its own and you want to see why the editor cut it that way. The agent also learns your format: show it a couple of reference reels and it carries your caption style and visual voice into the next edit. On pricing, Starter is $29 a month for 4 reels (about $7.25 a finished reel) and Pro is $59 a month for 12 reels (about $4.92 a reel), cheaper than a freelancer's fee to re-brief per clip.

Where we're weaker than the competition right now: we don't do automatic speaker separation yet (on the roadmap), so podcast captions aren't labeled per speaker. And we don't publish straight to platforms yet (the Reels, Shorts, and TikTok APIs); exports are MP4, and you drop them into a scheduler or the native app. For creators who live inside the TikTok mobile app, CapCut feels closer to home.

How they stack up

How to pick

• I already live in CapCut and Auto-Cut works for me → stay in CapCut. Any move costs time, and the gains from a different algorithm are usually smaller than the cost of losing the tool you already know.
• I'm a podcaster with one guest, one or two episodes a week, and I need clips weekly → Opus Clip. That's the core use case it's built for, and it does it better than anyone in the category.
• I'm in B2B marketing with a stream of Zoom calls and webinars → Vizard. The integrations pay back.
• I'm a coach, educator, or expert with mixed formats (lecture, talking-head, the occasional interview), and I want to see why the editor cut what it cut → Lotima. That's the contract we're building toward. For other cases we're not the best pick, and we won't pretend otherwise.

Why "best AI editor" is the wrong question

Creator communities keep asking which AI editor is best. There's no answer, because the question is put wrong. All four have their own contract: the source format they're for, how much they batch, how much control they give you. The tool that wins on a feature checklist will do worse on your format if your format sits outside what it's built for.

The better question is: which source format will fill my next three months, and which tool is built for it? Podcast, Opus. Zoom webinar, Vizard. TikTok-native creator, CapCut. Mixed formats where you need clips that carry a complete thought, Lotima, which is what we're building toward. If your format matches none of these, you probably don't need an AI editor yet, and manual cuts in Premiere will beat arguing with the algorithm.

In practice, spending the time to try all four and crown a winner only pays back if you're going to ship four or more reels a week for three months or longer. Under that, the choice between them barely matters; any of them saves you four to six hours a week over cutting by hand in DaVinci Resolve. The other three have free tiers that cover a test session; on Lotima the watermarked preview is free, you pay when you export, and a money-back window applies on top (see the refund rules). Either way, run one of your typical sessions through all four in an evening and decide on the finished file, not the copy on the landing page.

What to read next

• How to turn a 60-minute lecture into 6 reels: a workflow for educators
• Podcast to vertical reels: which moments actually work
• All articles