Runway Aleph review: an AI video editor, not another text-to-video model

Runway Aleph review from a motion designer: background swaps, object removal, relighting, new angles, what changed in Aleph 2.0, API prices and limits.
"Runway looks worse than Veo." I hear that a lot, and it's comparing apples to oranges. This Runway Aleph review is my attempt to explain why. I've been editing video by hand for about ten years, and these days I do motion design at Flami, where AI editors take roughly half of the routine off my plate. Runway is open on my screen pretty much every working day. It just does a different job from Veo and Kling, and judging it on their terms misses the point.
Let me show you what I mean.
So what is Aleph?
Runway calls it an in-context video model. You give it footage that already exists, and it changes that footage according to a text prompt. It doesn't start from a blank frame the way a classic text-to-video generator does.
The first version came out on July 25, 2025. Runway's announcement lists adding, removing and transforming objects. It also mentions new angles of a scene and changes to style or lighting. In plain English, the model looks at your clip, works out what's in it, and rebuilds the scene with new parameters instead of painting over the problem.
Here's a concrete case. Say you have a product clip on a white background, and you want the same product at sunset by the ocean. The old way went roughly like this:
- shoot on a green screen, or rotoscope the product by hand;
- comp it in After Effects;
- fix the light and shadows, because otherwise the fake shows;
- if the angle is wrong, you reshoot. Or you live with it.
With Aleph that's one prompt. Upload the clip, type "replace the background with a sunset over the ocean, keep the lighting on the product," and wait. In my experience about 6 attempts out of 10 are usable straight away. Three more need a second pass, and one is garbage.
Aleph 2.0 changed the numbers
On May 21, 2026, Runway shipped Aleph 2.0 together with a new Edit Studio. The pitch is precision:
"Aleph 2.0 makes targeted edits, changing only what you want to change and keeping everything else just as it was."
The practical differences matter more than the slogan. The product page talks about editing a single frame and having that look carried through the rest of the clip. You can also push one edit across several shots that sit on either side of a cut. According to the developer changelog, the API version takes input videos between 2 and 30 seconds with up to five keyframe images, and the original Aleph was retired from the API on July 30, 2026.
Most of my testing for this piece happened on the first Aleph. I haven't pushed 2.0 through the same set of jobs yet, so treat the hit rates below as first-Aleph numbers.
What it can do
Straight from Runway's descriptions of both versions:
- add objects to a scene;
- remove them;
- change how existing objects look, including product swaps;
- relight a shot or shift the time of day;
- restyle the whole frame;
- generate an alternative camera angle from a single shot.
On a traditional shoot, that's work for a small crew. A camera operator, an editor, somebody doing motion graphics, a colorist. Aleph covers each of those roles to a level that's good enough for most ad work.
Where it earned its keep in my tests
Background swaps for product video are the best feature, full stop, especially for Amazon listings and social ads. You shoot one clip of the product (or generate one in Kling), then make five versions with different backgrounds for different campaigns. Each version costs about the same as a single generation in a regular model.
Removing junk from the frame comes next. Someone films in a makeshift studio on a phone, and there's a cable on the floor, an outlet on the wall, a shopping bag in the corner. Aleph cleans that up in one pass. An editor used to spend an hour on it.
Changing the time of day was the one that made me grin. Product filmed in morning light, but the Black Friday campaign wants evening? Aleph relights the scene and stretches the shadows. It even warms up the color temperature. I honestly didn't expect it to handle shadow direction as well as it does.
New angles are the shakiest of the four. If you shot a scene from the front and need it from the side, Aleph can reconstruct the geometry and render the other view. It isn't perfect, especially on busy compositions. For a typical tabletop product shot it's probably good enough.
And where it doesn't
You can't make a video from nothing. Aleph needs source footage. If all you have is an idea and no clip, you want Veo or Kling instead.
Length used to be a real constraint. With the first Aleph I worked in short chunks, roughly five seconds at a time, and stitched the results. Aleph 2.0 accepts up to 30 seconds, which removes most of that pain, though a 30-second edit also costs six times as much as a 5-second one.
Complex animation and physics aren't its thing either. Editing an existing image is what it's built for. A dance or a jump or an explosion from scratch belongs to a generator.
Pure image quality is still Veo's, full stop, and no amount of Aleph tuning closes that gap. On a fragrance or jewelry shot Veo still has the edge, Aleph's edit reads as good, not luxe.
Blind voting agrees it's not the leader, at least for the old version. The arena.ai video edit leaderboard, last updated August 26, 2026, has runway-gen4-aleph in 10th place out of 10 with 1182, while Wan 3.0 and Seedance 2.5 sit on top. Aleph 2.0 isn't on that board yet. TechCrunch described it in July as ranking among the leading editing models, so I'd wait for the votes before writing it off.
What my week looks like now
Before AI editors, a typical ad went like this:
idea → brief to a studio → a week of waiting → first cut → notes → another week → done.
Now it usually looks like this:
idea → quick base clip in Kling → three or four edits in Aleph → done in about half an hour.
That's not a sales line, it's just how my calendar changed. Aleph removes the need for a studio and a separate editor on typical jobs. It's simply a much faster cycle than before.
Price and speed
Runway's API pricing lists Aleph 2 at 28 credits per second of video, with a 56-credit minimum per generation. Credits cost a cent each on the developer portal, so that's $0.28 a second, or $1.40 for a 5-second edit. Runway's own app plans count credits differently, so check there if you buy direct.
On Flami, Runway comes with the general subscription and is billed in credits like everything else. I'm not quoting a figure because pricing for international accounts isn't published yet.
A single edit took 30 to 90 seconds in my runs, which is faster than most video models. It makes sense, since the model is editing a scene that already exists instead of building one from zero.
Kling first, Aleph second
For product videos in bulk, this is the routine I actually use.
I start with the base clip in Kling 3.0, usually image-to-video from a product photo. Why Kling specifically is in my colleague's Kling 3.0 review.
When the client wants versions for different channels or campaigns, each copy of that clip goes through Aleph with a different prompt. New background on one. Evening light on another. A third gets a different angle, or an extra prop on the table.
So one base clip plus three variations costs a small fraction of a studio day. I'm running this on two client projects right now. The generators themselves are compared head to head in a separate piece.
Where Aleph fits among Runway's Gen models
It gets confusing here, because Runway ships models at a brisk pace. Here's how I keep them straight.
Gen-3 Alpha was the classic text-to-video and image-to-video generator. It's outdated now.
Gen-4 and its faster Turbo variant became the standard for generating from scratch. They handle text-to-video and image-to-video. Characters and locations stay consistent between shots. That's the one that competes with Veo and Kling.
Gen-4.5 arrived in December 2025, and shortly after that Runway added native audio and multi-shot generation to it, per TechCrunch's coverage. On the API it costs $0.12 a second against $0.05 for Gen-4 Turbo.
Aleph sits apart from all of them. It's the editing model, now in its 2.0 version.
On Flami's Runway page you'll find Gen-4 Turbo for generating and Aleph for editing. There's also Extend if a clip needs to run longer.
Should you use it?
Aleph earns its keep the moment you already have footage and want three or four versions of it: new backgrounds, a stray object gone, evening light instead of morning, a look for each A/B test. It also takes a real bite out of post-production costs.
Skip it if there's no source clip yet. That's a job for Gen-4 or Veo, maybe Kling. Same if your footage runs past 30 seconds uncut, you'll have to split it first.
Coming up
Next on my desk are Hailuo, Seedance 2.0 and Wan 2.7. I'm also putting together a walkthrough of building one ad through several models in a row.
That kind of stray clutter is exactly what an editor like this needs to prove itself on. Got the same clutter in your footage? Upload it to Flami and see how clean Runway gets it.
FAQ
What is Runway Aleph?
An in-context video editor from Runway. It takes existing footage and edits it from a text prompt. Typical edits are background swaps and relighting. It can also add or remove objects and render new camera angles.
Can Aleph generate video from scratch?
No. It needs a source clip. For generation from text you'd use Runway Gen-4 or Gen-4.5, or another model.
How long can the input clip be?
Aleph 2.0 takes 2 to 30 seconds per edit through the API. The first Aleph worked on much shorter chunks.
Aleph vs Veo 3.1, which is better?
Different question for different footage: got no clip yet, go to Veo. Already have one and want three versions of it, that's Aleph's job, and it's the one I reach for after the base clip exists.
Aleph vs Kling 3.0?
I start the base clip in Kling, then run three or four Aleph passes on top for backgrounds and light, usually under an hour for four versions.
How much does it cost?
On Runway's developer API, Aleph 2 is $0.28 per second with a two-second minimum. On Flami it's included in the subscription and paid in credits.
Do I need my own Runway account?
Not on Flami. Runway's models run inside the same subscription as Veo and Kling.
What resolution does it output?
1080p. Input footage can come in at different resolutions.
Last verified September 30, 2026, against Runway's documentation and the video edit leaderboard.
Sources
- Runway Research, Introducing Runway Aleph
- Runway News, Introducing Aleph 2.0 and Edit Studio
- Runway, Aleph 2.0 product page
- Runway Dev, API changelog
- Runway Dev, API pricing
- arena.ai, Video Edit Leaderboard
- TechCrunch, Runway bets on AI model routing
- TechCrunch, Runway adds native audio to its latest video model
- Flami, Runway page
About the author
Jake Turner
Reviewer at Flami