0 / 5000
Reference image to influence the style or content of the output.
Seed unlocked - will use random seed
AI Video Editor — Rewrite Footage, Keep the Motion
This AI video editor runs Runway's Aleph, an in-context video model that rewrites footage you already have: change the style, the lighting, the weather, even the camera angle — while the original motion carries through untouched. Upload an MP4 or WebM clip, describe the change in plain language, and collect the edited shot minutes later. No timeline, no keyframes, no plug-ins — and a candid guide below on what this kind of editing does brilliantly and where it stops.
Which Kind of Video Editor Are You Looking For?
Two very different tools share the name. Thirty seconds here saves you the wrong afternoon.
A timeline editor
Cutting clips together, trimming, adding music, captions, and transitions — the CapCut and Premiere category. Software arranges your footage; nothing inside the frame changes.
If that's the job, this page isn't your tool — any timeline app handles it better.
A prompt editor — this page
Changing what's inside the frame: the look, the light, the objects, the location — by describing the edit in words. The footage is re-rendered, not rearranged.
If the edit you imagine starts with "make it look like…", you are exactly where you should be.
The two combine well: rewrite the shot here, then cut, caption, and score it in your usual timeline.
Eight Edits Aleph Performs From a Sentence
Runway's documented capability set, annotated with how each one holds up in real footage.
Relight the scene
Swap professional lighting setups onto existing footage — golden hour, neon, overcast.
One of the most reliable edits; shadows track the original geometry convincingly.
Change season or weather
Summer to snow, clear to storm, day to dusk across the whole shot.
Strong on environments; double-check small background details on busy streets.
Generate a new camera angle
Produce a fresh viewpoint of the same scene from footage you already shot.
The most surprising capability — works best on scenes with clear spatial depth.
Swap objects or subjects
Replace a product, prop, or person while the action continues naturally.
Clean on a clearly framed subject; crowded frames raise the miss rate.
Remove what should not be there
Strip logos, bystanders, cars, or clutter without masks or rotoscoping.
Reviewers call it a genuine time-saver versus manual cleanup work.
Recolor with intent
Change the color of one object while lighting and reflections stay true.
Name the object precisely and it behaves; vague targets recolor too much.
Restyle the whole look
Carry a clip into anime, film noir, claymation, or a reference image's aesthetic.
Motion survives the transfer — that is the headline trick of in-context editing.
Age or de-age a subject
Shift a character's apparent age while performance and expression remain.
Effective in moderation; extreme jumps can drift identity.
Before You Prompt: What This Editor Can and Cannot Save
Aleph is editorial intelligence, not a repair shop. Set expectations here and the results follow.
It edits; it does not restore.
Broken focus, severe over-exposure, and chaotic framing stay broken — the model transforms what is legible, it does not reconstruct what is not. Shoot or pick the cleanest take you have.
Input quality is the output ceiling.
Stable, well-lit, clearly composed clips return the most convincing edits. A shaky phone clip will come back as a shaky edited clip — with better lighting.
Five seconds in, five seconds out.
Uploads can run longer, but only the first five seconds are processed — front-load the action you want edited, or trim before uploading. MP4 and WebM are accepted.
Plan for sound and minutes.
Output ships without native audio — score it in post. Renders typically land in two to ten minutes, longer at peak times; the queue is normal, not a failure.
The Model Behind the Tool
Runway Aleph
Runway
Runway has spent years building video models for film and ad workflows — its research line popularized text-driven VFX. Aleph is the company's in-context editor: instead of generating from nothing, it reads an input clip as a coherent, 3D-aware scene and applies your instruction to it.
Runway's own lineup makes the division of labor clear: Gen 4.5, its text-to-video flagship, invents footage from a prompt; Aleph rewrites footage that already exists. Different question, different model.
On this page Aleph does one job — prompt-driven video-to-video editing — with reference image support and a seed control for repeatable runs.
Three Workflows That Hold Up
Real edit patterns — with the prompt shape and the failure case for each.
One product shoot, every colorway
The footage: A five-second hero shot of the gray edition.
The instruction
"Change the jacket to burnt orange, keep everything else exactly as it is."
What returns: The same take in a new colorway — motion, light, and background untouched.
Why it holds: Recoloring a named object is among the most stable edits in the set.
Skip it when: Regulatory or print-accurate color matching — verify against a physical sample.
Reshoot the location, not the actor
The footage: A clean storefront walk-by filmed on an overcast Tuesday.
The instruction
"Set the scene at night in light rain, neon signage reflecting on wet pavement."
What returns: The same performance in a different world — schedule and weather stop being blockers.
Why it holds: Environment and lighting rewrites are Aleph's documented core strengths.
Skip it when: You need the new location to be a specific real place rendered accurately.
Stylize a sequence for a pitch
The footage: Rough live-action blocking captured on a phone.
The instruction
"Render in moody graphic-novel style, high-contrast ink shadows, keep the camera move."
What returns: A stylized previz that preserves your direction — enough to sell the idea before production.
Why it holds: Style transfer rides on the original motion, which is the in-context advantage.
Skip it when: Final-frame delivery at broadcast resolution — treat this as previz, not finishing.
Known Limits — and the Working Answers
Every one of these shows up in real reviews. None of them has to end a project.
Character details drift across separate renders — clothing, faces, small props.
Workaround: Edit a character's shots in as few passes as possible, attach a reference image, and lock the seed when you need the same result twice.
Heavily damaged source footage comes back damaged.
Workaround: Fix exposure and stabilization in a conventional tool first, then send the corrected clip in for the creative rewrite.
The five-second window cuts longer stories short.
Workaround: Break the sequence into five-second beats, edit each with identical style wording, and assemble on a timeline.
No native audio on output.
Workaround: Keep the original track if the action timing matches, or rebuild sound in post — and write the edit so timing-critical action stays put.
Render times stretch at peak hours.
Workaround: Batch your prompts: queue several variants in one sitting and review them together instead of waiting on each run.
Prompting Aleph: Say What Changes, Pin What Stays
Distilled from Runway's official Aleph prompting guide, then checked against field results.
The two-part instruction
Aleph prompts differ from generation prompts: you are not describing a scene, you are describing a delta. State the change precisely, then pin what must not move — subject, motion, framing. Runway's guide recommends exactly this change-plus-preserve structure.
"Turn the daylight street into a rainy night with neon reflections — keep the walking pace, the camera path, and the subject's red coat unchanged"
The same edit, asked twice
Vague
"make this video look cinematic and cool, neon style"
Pinned
"Apply a neon-noir grade: cyan-magenta palette, wet asphalt reflections — preserve the original camera dolly and the subject's features"
"Cinematic" hands the model every decision at once. The rewrite names the palette, adds one concrete surface cue, and explicitly protects the two things edits most often break: the camera move and the face.
Field grammar that keeps working
- Name targets like a colorist: "the cyclist's jacket", not "the clothes."
- One transformation per render — stacked changes multiply drift; chain renders instead.
- Reference images outrank adjectives for style: attach the look you mean.
- Re-running a keeper? Lock the seed and reuse the exact wording.
Aleph, a Timeline, or a Fresh Generation?
Three tools, one decision rule each.
Rewrite it here
The footage exists and the change is inside the frame — look, light, objects, environment, angle. The identity of the take matters.
Cut it in a timeline
The change is between frames — order, pace, music, captions, transitions. Nothing inside the shot needs to differ.
Generate from scratch
The shot you need was never filmed. A text-to-video model invents it — at the cost of starting identity from zero.
How the AI Video Editor Works Here
Three moves from clip to rewritten clip — the tool sits at the top of this page.
Upload the clip
MP4 or WebM. The first five seconds are the canvas, so lead with the action — trim beforehand if the moment sits later in the file.
Describe the delta
One change, precisely named, plus what must stay. Add a reference image when a style is easier to show than to say; set the aspect ratio to match the destination.
Collect and iterate
Renders land in roughly two to ten minutes. Review at full size, adjust one variable, and lock the seed once a take is worth repeating.
AI Video Editor: Straight Answers
Scoping questions first, craft questions second — answered from official docs and field reports.
Where the Clip Goes Next
Generate new footage, transfer a performance, or build the stills around it.
Your Footage Already Has a Second Draft
Restyle it, relight it, strip the clutter, move it to midnight rain — while the camera move and the performance stay yours. Upload a clip and ask this AI video editor for the version you actually needed.