AI Video Editor: Describe the Edit, Approve the Plan
BetaThis AI video editor turns a sentence into an edit. Type what you want, such as “cut the pauses, add boxed captions and make it vertical for TikTok”, and ProMedia proposes a plan built from its own tools. You approve, reorder or drop steps, then the plan runs in your browser on your video. It is free, in Beta, and nothing is uploaded.
- Runs on your device
- Whisper AI
- On-device AI
- Works offline
Updated
This tool is in beta: it works, but it is heavy for some devices and results may need a second look.
How to edit a video with AI
Step 1: Drop one video on the box above: a talking-head clip, a tutorial, an interview or a podcast recording.
Step 2: Describe the edit in plain words under Describe your edit, or tap a suggestion such as Cut pauses & ums or Vertical for TikTok.
Step 3: Press Plan the edit. Check the steps: switch any off, move them up or down, or add one from Add a step.
Step 4: Press Run to apply the plan on your device, then download the edited video and any extra results.
Good to know
Formats
How it runs
- Whisper speech model
- Speech recognition uses an open Whisper model. It downloads once after you confirm its size, then runs on your device — on the GPU via WebGPU when available.
- Vision model
- A compact computer-vision model runs locally in your browser. Nothing is sent to an AI service.
- Hardware video engine
- Video is decoded and encoded by your device’s media hardware through WebCodecs — usually several times faster than real time.
What you can ask the AI video editor to do
Each part of your sentence becomes one step that uses a real ProMedia tool with matching settings. These requests were run through the planner word for word; the right-hand column is the plan it produced.
| You type | Plan |
|---|---|
| cut the pauses, add boxed captions and make it vertical for TikTok | Cut pauses (fillers off) → Make it vertical (9:16) → Burn in captions (Boxed) |
| remove the ums and bleep the swear words | Cut filler words (pauses off) → Bleep swear words |
| clean up the audio and stabilize it | Reduce background noise → Stabilise shaky footage |
| speed it up 1.5x and compress it under 25 MB | Change speed (1.5×) → Compress (under 25 MB) |
| add karaoke captions at the top | Burn in captions (Karaoke, top) |
| blur the faces and mute it | Blur faces → Remove the sound |
| make it square for instagram | Make it vertical (1:1) |
| write chapters and a YouTube description | YouTube chapters + title, description and hashtags (text results) |
How the AI video editor works
Planning and editing are separate on purpose, so nothing happens to your video until you have seen the plan.
- You describe the edit. Up to 500 characters of plain English; commas, “and”, “then” and “also” separate the steps.
- A planner maps it to tools. In desktop Chrome with built-in AI (Gemini Nano), an on-device language model picks the tools. Everywhere else a rule-based planner does it. Neither sends your words anywhere.
- You approve the plan. Every step can be switched off, moved, opened in its own tool page, or added from a list.
- The steps run in order on your device. Each step is the same processor its own tool page uses; the edited video is handed to the next step, and transcripts, chapters, MP3 audio or GIFs come back as extra downloads.
Why the steps run in a fixed order
Some edits only make sense before others. Pauses are cut before captions are made, so the captions match the shorter video; compression comes last, so nothing re-encodes a file that was already squeezed. The planner sorts every plan into this order, and you can still move steps by hand.
| Order | Steps | Why here |
|---|---|---|
| 1 | Cut pauses and filler words, bleep swear words | Changes the timing everything later depends on |
| 2 | Reduce background noise, stabilise, change speed | Fixes the sound and picture before they are framed |
| 3 | Colour filter, blur faces | Works on the full frame |
| 4 | Reframe around the speaker, make it vertical, flip, rotate | Sets the final shape |
| 5 | Burn in captions | Drawn on the final shape, at the final timing |
| 6 | Remove the sound, compress (for Discord, WhatsApp, email or a size) | Last, so the file is encoded once more at most |
| 7 | Transcript, YouTube chapters, title and description, MP3, GIF | Extra results; they do not change the video |
What is text-based video editing?
Text-based video editing means changing a video by working with words instead of a timeline. In editors that popularised it, you delete words from the transcript and the matching video disappears. ProMedia’s editor takes the other text route: you write the edit itself as a sentence and get a plan of tools.
The closest thing to transcript editing here is the filler-word step: it removes “um”, “uh” and set-off “like” from the video using word timings, without you touching a timeline. For word-by-word control over captions, the subtitle generator has an editor for every word.
Is an AI video editor good enough for YouTube?
For the repetitive part of editing, yes. Tightening a talking-head video with remove silence from video, adding captions, reframing for Shorts and compressing for upload are mechanical jobs, and the plan does them in one run. Storytelling is still yours: choosing takes, adding B-roll, music and transitions needs a timeline editor.
| Request | Supported? | Instead |
|---|---|---|
| Cut pauses, fillers, swear words | Yes | — |
| Captions, vertical or square framing, face blur | Yes | — |
| Noise reduction, stabilising, speed, colour filters | Yes | — |
| Compress to a size or for Discord, WhatsApp, email | Yes | — |
| Trim to a time (“cut the first 10 seconds”) | Not understood yet | Use the video trimmer |
| Add music or a logo | Not understood yet | Use add music to video |
| Combine several clips | No: one video per plan | Join them first, then edit |
| Generate new footage or images | No | It edits the video you give it |
Free, private and in your browser
- No upload. The plan and every step run on your device; there is no server that receives video.
- No watermark or credits. Run as many plans as you like.
- Models download once. Steps that need a speech model ask before its first download, then work offline.
- Beta, honestly. Free-form planning is new, so check the plan before you press Run; a step that finds nothing to do, such as a pause cut on a video without pauses, is skipped and noted.
Which browsers and devices work
The editor runs in current desktop and mobile browsers with the rule-based planner. Planning with on-device AI needs desktop Chrome with its built-in language model; when it is ready, a switch called Plan with Chrome’s on-device AI appears. Long videos and several speech steps go faster on a laptop or desktop, especially one with WebGPU.
Tips & troubleshooting
Be literal: “cut the pauses and add captions” plans better than “make it more engaging”.
Put the platform in the sentence. “vertical for TikTok” sets 9:16, and “under 25 MB” sets the compressor’s size target.
If a step needs fine control, use the arrow button to open it in its own tool, where every setting is available.
Turning a long recording into several Shorts is a different job: the AI clip maker finds the best moments for you.
AI Video Editor (Describe an Edit) FAQ
How do I edit a video with AI?
Drop the video here, describe the edit in a sentence, press Plan the edit and check the proposed steps. Then press Run: each step, such as cutting pauses, adding captions or making it vertical, runs on your device and the edited video is ready to download.
Is the AI video editor free?
Yes. There is no subscription, credit system, export limit or watermark. Everything runs in your browser, and the only downloads are the engines and AI models your steps need, each fetched once; AI models ask first.
Can I edit a video with an AI prompt?
Yes, within the editing jobs ProMedia has tools for: cutting pauses and filler words, bleeping, noise reduction, stabilising, speed, filters, face blur, reframing, captions, muting and compression, plus transcripts, chapters, descriptions, MP3 and GIF. Requests outside that list are reported as not understood.
Does any step run in the cloud?
No. Planning happens in your browser, with the rule-based planner or Chrome’s on-device model, and every editing step runs on your device. Your video and your prompt are never uploaded.
Why is the AI video editor in Beta?
Free-form requests can be phrased in endless ways, and the planner understands the common ones, not all of them. That is why you approve the plan first. If a request is not understood, rephrase it or add the step yourself from Add a step.
Can it make a video from text?
No. It is an editor, not a video generator: it changes the video you give it. To turn text into a voice-over for your video, use text to speech and add the audio with add music to video.