Need to publish a horizontal clip vertically? Send it to the bot, pick a format — and get the finished file back. Flowmark looks at what is happening in the frame: it holds on to faces, and when the main thing isn’t a person — an object, an animal, an event — it notices that too. On its own, or from your description.
Promo access opens every Flowmark tool for 30 days — auto-crop, branding, compression, transcription and more. Right in Telegram and MAX.
the man in the blue jacket on the right
Footage is almost always horizontal, while publishing is increasingly vertical: Reels, Shorts, TikTok, Stories. Between the two sits the routine of cropping, which usually means downloading the file, opening an editor, manually dragging the frame after the speaker, rendering and uploading again.
Auto-crop removes that loop: the video or photo is already in Telegram — and it gets re-cropped right there. Send the file, pick the format, get the result in the same chat.
The tool used to work mostly from faces. Now it also notices other things that matter — an object, an animal, an unfolding event — on its own or from your description.
“Automatic” — the bot decides for itself what matters in the frame. “People and faces” — it stays with the people and ignores other eye-catching objects. “Name the subject” — you say in words what to keep.
“A burning car and smoke”, “the man in the blue jacket at the top” — the bot separates the object from the action and looks for exactly that. Hints about position work too: “on the left”, “top right”, “in the bottom left corner”. If it can’t find it, it says so and falls back to automatic cropping.
The clip is split into scenes and cropped scene by scene. Inside a scene the window doesn’t chase the subject: while the main thing sits safely in view the frame stays put, and if the subject does drift away the window travels smoothly rather than snapping.
1:1 — square for the feed, 9:16 — Reels, Shorts and Stories, 16:9 — horizontal, 3:4 and 4:3 — for posts and covers. With video you can ask to remove black bars if the clip has already survived someone else’s cropping; with photos, white and black bars are detected and removed on their own before the bot looks for the subject.
The bitrate is recalculated for the new frame area, the frame rate is preserved, the audio is not re-encoded. Photos stay in their original format — a transparent PNG remains a PNG.
Finding an object or an event already works in real jobs, but a difficult shot can still throw it off. When the bot isn’t sure, it doesn’t move the frame on a guess — it falls back to the proven cropping by faces and composition.
The re-cropped clip can be branded with a logo, compressed to a platform limit or captioned right away — the whole media workflow in one bot:
A center crop does not know what is happening in the shot and often cuts off the most important part. The bot looks at the frame first: it finds faces, and where needed an object, an animal or an event, and fits the crop window so the main thing stays visible.
The clip is automatically split into scenes, and the crop window is chosen anew for each one — from whatever is happening in that particular scene.
In an ordinary phrase: “a burning car and smoke”, “the man in the blue jacket at the top”, “the dog by the door”. The bot separates the object from the action and looks for exactly that. If it cannot find it with confidence, it says so and does ordinary automatic cropping instead of guessing.
If all faces physically cannot fit into the chosen format, the frame is centered on their weighted center. You get the result right away and see whether the trade-off works.
Only when there is no way around it. While the main thing stays safely in view the window does not move — a floating frame looks automated and distracts from the content. If the subject does drift towards the edge, the frame travels there smoothly instead of snapping after it.
Yes, with the same tool and the same modes: the photo is cropped to the chosen format and saved in the original file format.
Finding an object or an event is the newest part of the tool: it is already useful in real work, but a difficult shot can still throw it off. When the bot is not sure, it does not move the frame on a guess — it falls back to the proven cropping by faces and composition.
There's a free plan, and the promo link opens 30 days of extended access to every tool.
