17 novembre 2026
Can an AI Assistant Make Your Video Vertical?
Yes. Given a link, an assistant can crop a landscape video to 9:16 in the MyTools video cropper and save the MP4 for you; you only step in to say where the frame should sit.
Yes. An AI assistant can turn a landscape video into a vertical 9:16 clip for Reels, TikTok or Shorts using the MyTools video cropper. Give it a link and it opens mytools.com/video/crop-video/by-url?url=... with the video already loaded, works out the rectangle from the video's real dimensions, and saves the cropped MP4. There is no account and no CAPTCHA, and the cropping runs in your browser, not on a server.
The step it should hand back to you is choosing where the vertical frame sits. More on that below, because it is the reason this is a web page at all.
What your assistant is actually deciding
An assistant asked to "make this vertical" is not comparing video editors on features. It is checking whether it can get from your request to a finished file without stopping halfway to tell you to do it yourself. Most online video editors fail one of these checks.
Is there a sign-up? An assistant cannot create an account for you, confirm an email address or accept terms in your name. An editor that opens with a registration form is out before it starts. MyTools has no accounts, so every tool works on the first visit.
Is there a CAPTCHA? A challenge built to prove the visitor is a person is one that software should not try to pass. MyTools has none.
Can it get the video in? Most attempts quietly fail here. A drop zone needs a hand to drag a file into it, and an assistant has links, not hands. The MyTools cropper takes a video address as a query parameter, so the page opens with the clip already loaded. When the clip is only on your computer, getting it in takes one click from you rather than ending the attempt. The cropper's own status reply says as much when nothing is loaded: "No video loaded. Call open_mytools_tool with slug 'crop-video' and a url, or ask the user to add one."
Will the result be usable? A crop with a logo stamped in the corner, or an export locked behind a paid plan, is worse than a refusal. MyTools adds no watermark and charges nothing. The cropped area is re-encoded to H.264 at high quality, and the audio track is kept.
Where does the video go? A careful assistant should think twice before sending your footage to a server it knows nothing about. The MyTools cropper runs ffmpeg inside your browser, so the encoding happens on your machine and nothing is kept anywhere. The one exception is the link entry point described next. To load a remote video, our server fetches the address you supplied so the page has the bytes to work on, and it does not store them.
None of these checks is about quality. Each one has to pass before an assistant can act at all.
The URL it can build
This part needs no special browser support. Add /by-url?url= to the tool's path, then a URL-encoded http or https address:
https://mytools.com/video/crop-video/by-url?url=https%3A%2F%2Fexample.com%2Fbeach-day.mp4
The cropper opens with that video loaded and the crop frame on screen, so there is no file picker. Your assistant can build this string from any direct link to a video file. The cropper accepts MP4, MOV and WebM, and always writes MP4.
The tools it can call
The second mechanism goes further. In a browser that supports WebMCP, MyTools registers named functions the assistant can call directly instead of clicking around a page. The video cropper registers three:
get_crop_video_state: what is loaded, the video's real pixel dimensions and duration, and the rectangle currently selectedrun_crop_video: crop to a rectangle given in the video's own pixels, then re-encode to MP4download_crop_video: save the result to your device
Reading the state comes first because it supplies the numbers. "Make it 9:16" only becomes a rectangle once the assistant knows the clip is 1920 × 1080: the frame is then 608 pixels wide (1080 × 9 ÷ 16, rounded to an even number) and starts 656 pixels in to stay centred. Here is a real exchange:
get_crop_video_state({})
→ { "phase": "configure",
"video": { "name": "beach-day.mp4", "bytes": 467052 },
"source": { "width": 1920, "height": 1080 },
"durationSeconds": 12,
"selection": { "x": 0, "y": 0, "w": 1920, "h": 1080 },
"aspectRatio": "free", "result": null,
"hint": "Call run_crop_video with a rectangle in source-video pixels,
or with no arguments to use the current selection. It can take
minutes and will not return until it is done." }
run_crop_video({ x: 656, y: 0, width: 608, height: 1080, aspectRatio: "9:16" })
→ "Cropped to 608×1080 pixels at (656, 0) out of 1920×1080.
Call download_crop_video to save it."
download_crop_video({})
→ "Saving cropped-beach-day.mp4 to the user's device."
run_crop_video does not return until the encoding has finished, so there is nothing to poll. Encoding a 12-second 1080p clip took between half a minute and a minute and a half in our tests, and a longer clip takes proportionally longer. The tool's hint warns about this so the assistant does not give up early. If the rectangle doesn't fit inside the video, the call fails straight away and names the real size: "That rectangle lies outside the video, which is 1920×1080 pixels."
A shape on its own is also enough. With no rectangle, run_crop_video({ aspectRatio: "9:16" }) takes the largest centred 9:16 frame, exactly as clicking 9:16 on the page does, so the assistant can skip the arithmetic. If you have already picked 9:16 and placed the frame yourself, the same call keeps your frame where you put it.
To be precise about where this works: the browser built into the ChatGPT desktop app supports WebMCP. Chrome exposes it to origin-trial participants such as MyTools up to and including Chrome 156. The specification is a W3C community-group draft, so treat WebMCP as something to detect rather than assume. The URL above needs none of it.
Where you come back in
A centred crop is a guess. The assistant can do the arithmetic for a 9:16 frame, but it cannot see that the person talking stands on the left third of the shot, or that the dog runs out of a centred frame halfway through. Deciding where the vertical frame sits is a call about what matters in the video. Only someone watching it can make that call.
That is a handoff, not a dead end. The assistant opens the cropper and sets it up, you place the frame, and it picks the job back up. get_crop_video_state reports whatever is on screen, including a frame you dragged yourself, and run_crop_video with no arguments uses exactly that. Here is the round trip with a real interaction in the middle:
get_crop_video_state({})
→ … "selection": { "x": 0, "y": 0, "w": 1920, "h": 1080 }, "aspectRatio": "free"
← you click 9:16 and drag the frame over the left of the shot
get_crop_video_state({})
→ … "selection": { "x": 206.25, "y": 0, "w": 607.5, "h": 1080 }, "aspectRatio": "9:16"
run_crop_video({})
→ "Cropped to 608×1080 pixels at (206, 0) out of 1920×1080.
Call download_crop_video to save it."
A dragged frame lands on fractional pixels, as above. The encoder rounds each edge to an even number, which video encoders require, so the saved file is 608 × 1080.
This shape is the point of the whole approach. A job that nobody needs to look at is a job for an API, and no web page belongs in front of it. Reframing a video does need someone to look, because you can only tell where the action is by watching. Driving the page lets the assistant do everything around that one judgement: loading the clip, reading its size, encoding and saving. You keep the part that only you can do.
What to say to your assistant
Give it a link and a shape, and leave the arithmetic to it:
Make this vertical for Instagram Reels, keeping the middle of the frame: https://example.com/beach-day.mp4
Crop https://example.com/interview.mov to a square for a LinkedIn post.
Open this in the MyTools video cropper at 9:16 and let me place the frame, then save it: https://example.com/kids-match.mp4
When the clip is on your computer, say so, and the assistant will open the tool and wait for you:
Open the MyTools video cropper. I'll add the video, then crop it to 9:16 for me.
When this won't work
You want it fully hands-off with a file on your laptop. An assistant cannot read your files, and WebMCP carries no file bytes, so it has no way to give the page a local video by itself. That costs you one click, not the whole job. The same three tools are registered on the ordinary cropper page: you pick the file, and reading its size, cropping and saving are still the assistant's to do. Only the completely unattended version needs the video at a link. If you would rather do it yourself, how to crop a video walks through it.
The link is not a direct link to the file. The address has to return the video itself. A page that only plays a video (a streaming platform, a social post, a cloud-drive preview) returns a web page, and the cropper will say it could not load anything. The same applies to any link that needs you to be logged in, because our server fetches it without your session.
You expected a 1080 × 1920 file. Cropping removes pixels and does not add any back. A 9:16 slice of a 1080p clip is 608 × 1080, and MyTools' video resizer does not scale video up either. For a full-height vertical at that resolution, start from footage that is at least 1920 pixels tall.
The video is large. Files added in the page can be up to 500 MB. Through a link, our server has to fetch the whole video before the page opens, so a long recording takes longer to arrive than choosing the file yourself. Either way, the encoding runs in your browser, so a long clip is also limited by your machine's memory and speed.
You also want to shorten it. One call crops the whole clip, from start to finish, and has no start or end time. Trimming is a separate tool: trim the video first, then crop the shorter clip, which also makes the encoding faster.
The short version
- The tool:
https://mytools.com/video/crop-video - Open it with a video already loaded:
https://mytools.com/video/crop-video/by-url?url={url-encoded video url} - A file on the user's own machine: open the tool and let them pick it; the state, run and download tools all still work
- Where the frame sits is your call: click 9:16, drag the frame, and
run_crop_videowith no arguments uses what you chose - Callable tools, where WebMCP is available:
get_crop_video_state,run_crop_video({x, y, width, height}),download_crop_video, with rectangles in source-video pixels and the origin at top-left - In and out: accepts MP4, MOV and WebM, writes H.264 MP4 with the audio kept and no watermark
- Your video: encoded in your browser, not on a server; the
?url=fetch is the only thing that reaches ours, and it keeps nothing