Make a Short Video From Your Agent: Motion Design or an AI Creator, in One Call
A launch, a price change, a new integration: the post goes out faster when it comes with a video. Until today your agent could write the post but not the video. Now it can.
create_video is live on Cogny's MCP endpoint. Give your agent the lines a short video should say and it comes back with a finished MP4 in your Creative library. There are two styles:
- Motion design: animated type in your logo and brand colours.
- Influencer: an AI-generated creator says your lines to camera.
TL;DR
- One tool, two styles.
style: "motion_design"(the default) orstyle: "influencer". - Your words, exactly. 1 to 6 lines, used word for word. Nothing is added.
- Motion design checks itself. Cogny previews the frames, fixes what a layout check and a visual review flag (up to three previews), then renders. Every line stays on screen long enough to read, and
emphasismakes one line the hero: biggest type, held longest. - Influencer speech is checked. Cogny transcribes the clip and compares it with your script. Every number has to be right. If it isn't, the clip is made once more; if the second take misses too, you get the better one with a warning that quotes what was said.
- Formats: 4:5 feed (1080×1350), square, landscape (1920×1080) and vertical (1080×1920) for Stories, Reels, TikTok and Shorts. Influencer is vertical or landscape.
- Cost: motion design about 2 to 4 credits. Influencer about 1.3 credits a second (10 seconds is about 13 credits), and twice that at most if the speech has to be redone.
- Every plan, paid from credits.
Motion design: your lines, set in your brand
The first line is the headline and the last is the call to action. Cogny writes the animation in your workspace's brand colours, with your logo from Brand & Context (or a brand asset you pick). It renders frames and checks them before making the video:
- Layout: text inside the frame, nothing overlapping, nothing cut off.
- Readability: each line holds for about a second plus 0.3 seconds a word.
- A visual review: a second look at the frames that flags anything that reads badly, including "another line is bigger than the hero".
If something is off, it fixes it and previews again, up to three times, then renders. It usually takes one to three minutes.
Name the line that matters with emphasis: the price, the punchline, the date. That line is set biggest and stays on screen longest.
Influencer: an AI creator says your script
For the feel of a creator talking to a phone camera, use style: "influencer". Describe who they are and where in concept ("late twenties, at home by a window, telling a friend about a tool"). The creator is always AI-generated, never a real, named or lookalike person.
What the creator says is your script, said properly. Prices, numbers and web addresses are spoken in full, so "$280 a month" is said "two hundred and eighty dollars a month" and "cogny.com" "cogny dot com". The clip's length is planned on those spoken words: about 2.5 words a second, 3 to 10 seconds.
Then Cogny listens back. When the clip arrives, Cogny transcribes its audio without seeing the script and compares the two. Every number has to match. The other words may differ a little, for example a brand name spelled differently by the transcription. If it doesn't match, the clip is made once more with a slower, clearer delivery. If the second take doesn't match either, you still get the better of the two, flagged, with the line the creator actually said. The result tells your agent whether the speech was checked and what it found.
The clip has no text in the frame. Use your lines as written for the post's caption.
What it costs
| Style | Cost | Charged when |
|---|---|---|
| Motion design | About 2 to 4 credits (about 1 per frame preview and the render) | As each preview and the render are made |
| Influencer | About 1.3 credits a second: 8 s ≈ 11, 10 s ≈ 13 | When the clip is delivered. A redo is charged too, so at most twice that |
On Cogny AI, Pro and Cloud a video is paid from your monthly credits first. On Free and Solo it is paid from top-up credits. Your agent confirms the cost with you before it starts, and a workspace without enough credits gets a link to top up instead. You can stop a video with cancel_video; you pay only for what was already made.
Try it
From Claude, ChatGPT, Cursor or Claude Code connected to app.cogny.com/mcp:
"Make a 4:5 motion design video with these lines: 'Import from Vercel is live' / 'Bring your site over with one token' / 'cogny.com'. Make the first line the hero."
"Make a 10-second vertical influencer video where a creator says: 'I moved my site to Cogny with one token' / 'Now my AI edits it for me' / 'cogny.com'."
Your agent follows the video with get_video and gets the MP4 link, and the video is already in the Creative library. The full tool reference is in the MCP docs.