Flowmark is a practical tool for authors, editors and teams who need to process a steady flow of photos and videos quickly and consistently. Logos, watermarks, face masking, smart cropping, video editing, compression, audio-to-text, and profanity muting — without leaving Telegram or MAX.
Extended access opens more capabilities for 30 days, so you can test Flowmark on real files. After that, Flowmark remains available for free.
Flowmark Dossier 001
EDIT · 05/09/2026
Flowmark is a calm working layer for media processing: branding, captions, face masking, compression and speech tools, right where conversation and production already happen.
Flowmark does not try to impress with effects. It handles repeatable media operations: add a logo, add a watermark, add a caption, mask faces, re-crop a frame to another format, compress video to a target size, transcribe speech, or mute unwanted audio — quickly, consistently, and without unnecessary switching between tools.
The bot works inside Telegram and MAX, and the settings live in the Mini App — a built-in panel inside the messenger. Only quick actions remain in the chat: send a file, choose the right button, get the result. In the Mini App you can configure assets, logo versions, presets, captions, face masking, video parameters and button layout.
The main principle is simple: frequent actions should not lead a person away every time from chat to a separate program. File → button → finished file.
The logo or watermark can be placed in one of nine positions, rotated to any angle, repeated as tiles across the frame, or stretched to fill the entire screen. Its size, transparency, outline, and noise protection can be adjusted for different tasks. A logo can also be a short video with a transparent background — an animated emblem is overlaid on clips just like a regular image, and on video the logo can gently sway or change position. The same asset can have several versions and switch directly from the bot. Several logos and captions can be combined on one file as layers — and arranged not only with buttons, but with a phrase in the chat: “logo 1, opacity 20 + logo 2 top left”. It works for a batch of files too.
Videos can be quickly prepared for sending, autoplay, or fitting the limits of a specific platform: compress by preset, fit to a given size, remove sound or convert GIF and video. A separate button turns a clip into a round video message, which can be given a look of its own — from a colour frame to Matrix. The system itself selects a balance between size, speed and quality, so that you don’t have to manually select export parameters each time.
Trim a clip or join several into one without opening an editor: send the video and write in plain words what to do with it — “remove 2:30 to 2:48”, “drop the first thirty seconds”, “third one first, then the first”. Timecodes are understood in almost any notation, and you can refer to the words spoken in the clip instead. Before assembling, the bot shows you in the chat what you are going to get: which parts remain and where the cuts land — and you can correct it with the next message. The tool is still in beta.
The frame is chosen the same way: auto-crop reworks a horizontal clip or photo into vertical and square formats, keeping what matters in view — faces, an object or the action, on its own or from your description. Frames for round video messages are prepared on the same principle.
Flowmark can turn speech from audio or video into text: with timecodes, speaker separation and different output formats; the language can be set explicitly, and the formatting of the result can be changed after it arrives. A separate mode mutes obscene words in audio or video: with a custom group dictionary, a review of every trigger and a free rebuild. Voice is separated from music and background noise automatically — there is no mode to pick — and Russian speech is double-checked by a second recognizer tuned for the language. For video editing work, you can get not only a processed file, but also XML with intervals for manual editing.
Flowmark finds faces in an image and lets you mask them all at once or customize each one separately. Pixelation, blur, and a solid color mask are available; you can cover the whole face or just the eye area. For complex shots, there is manual adjustment of the area and strength of the effect.
Auxiliary neural-network tools. The AI editor changes photos from a text request — “remove the extra object”, “make the background wintery” — and keeps the whole chain of edits visible in the chat: a panel under the latest result lists the original photo and every variant, and the bot says which one your next request will change. Up to five images can be sent as input. A picture can also be created from scratch: describe it to the bot in an ordinary sentence — it works out that you want a generation and offers a button. The logo generator builds a mark with a transparent background — a quick start for those who do not have a logo yet. Every request passes pre-moderation before launch.
The caption editor lets you prepare the design once: grid position, margins, font, color, background, corner radius, line spacing and an icon from group assets. You can create as many presets as needed; in the bot they stay as quick buttons, while the Mini App shows a live preview.
A team in Flowmark is a group: a shared workspace with a shared brand kit for everyone. It stores assets, versions, presets, captions, button layout and tool settings. Each participant has their own role (owner, administrator, participant), but the bot itself remains the same for everyone: the same buttons, the same result. Participants can work from Telegram and MAX: the invitation is sent in one messenger, but you can receive it in another.
Several files can be processed in a batch — automatically, through manual batch mode or as an album. The group configuration can be uploaded to .flowpack: make a backup, transfer the settings to another account or transfer to another team.
The owner and administrators see team usage: how many credits went out over a period, how the spending is split across participants and tools, and a feed of operations. Each participant checks their own spending in the “Profile” section.
Flowmark does not force you to go through the same path every time. The bot detects the file type and shows the right buttons: for video — compression, cropping, branding, and sound; for audio — transcription and profanity muting; for images — captions, assets, face masking, and the AI editor. Which buttons show up under a file is decided not in the chat but in the mini app: the neighbouring tab above the mockup.
The bot does the work; the mini app decides how exactly it is done: which logo goes on, how a caption looks, which buttons show up under a file at all. It opens right inside the messenger — from the bot menu, with no installation and no separate account. Changes take effect on the next file you send, so you set everything up once and come back only when something actually needs to change.
You can try it on your file: open the bot, send a photo, video, or audio file and select the desired action.
Flowmark does not store your original photos, videos, and audio files permanently: they are used for processing and deleted after sending the result, within a maximum of 24 hours.
This is the normal working path for tasks that require manual assembly, precise assembly or non-standard editing. Flowmark does not replace editors and editing programs - it takes away repeatable operations that don't have to start from the timeline every time.
For the editor, it is a way to avoid waiting on separate processing and not waste time on the same type of export when the task is already clear.
For one frame the path is tolerable. When there are dozens of pictures a day, each one is repetition: the same ten minutes, multiplied by the number of files and people.
The album is processed in the same way: several photos are processed in a batch and are returned with the same design.
Since the public launch, Flowmark has processed 5,274 operations in real work chats. Not demo files and not a synthetic test: this is what users and teams send to the bot on an ordinary working day.
Every render the user asked for is counted, including free touch-ups of an already finished result. The numbers are refreshed when the site is published.
Before the public launch, Flowmark existed for several years as a closed internal version for one work team with a regular media stream. Below are its statistics: the same basic scenario, the same “file → button → result” principle, the same operations that later became the basis of the service.
The main scenario of the closed version was fast branding of the stream; everything else was added as that route picked up neighbouring tasks. These numbers were frozen at the public launch and no longer change.
The source lands in temporary storage — only for the duration of processing, nothing else.
Logo, caption, crop or transcription — the task runs in the group's isolated queue.
Right after the result is delivered, the source and working copies are removed from the server.
An automatic mechanism destroys everything older than a day — including interrupted and abandoned tasks.
This is how the system works, not a policy clause: the mechanism operates at the infrastructure level and does not rely on manual procedures. Only brand kit assets, group settings and operation history are stored permanently — the things the user uploaded and configured themselves. Details are in the privacy policy.