Motion Design Videos From Any Cogny Agent: Launch Videos, Product Updates and Ads, Rendered to MP4

Short videos with text on screen are the default format for launches, product updates and feed ads. Until now a Cogny agent could write the copy and generate an image, but the video still meant opening an editor.

That step is gone. Any Cogny agent can now make a motion design video: animated type, your logo, your colours, the exact words you approved. It checks its own frames before rendering, and the MP4 lands in your Creative library ready to post.

---

TL;DR

• Two new tools on the Image Gen & Media server: preview_video_frames and render_video. They work in Cogny chat, in loops and tickets, and from any MCP client connected to Cogny.
• Formats: 4:5 feed (1080×1350, the default), 1:1 (1080×1080), 16:9 (1920×1080) and 9:16 (1080×1920). Up to 60 seconds, 30 fps, silent.
• The agent checks its work. Every preview returns a sheet of frames, layout problems such as overflowing or overlapping text, and a short visual review. The agent fixes those and previews again before it renders.
• The result is an MP4 with a public link, saved to Creative. In Cogny chat it plays inline.
• Posting: attach it to an X post (upload_media + create_post) or a LinkedIn post (create_post with video_url).
• Cost: in practice 1 credit per preview and 1 per render. A video that fails the checks is sent back with the errors to fix and isn't charged.

---

Motion design or AI video?

Cogny now makes two kinds of video, and they're good at different things.

If the words on screen matter, use motion design.

---

How it works

The agent writes the video as one HTML file: the scenes, the type and the timing of every fade and move. That file is the video, so the agent can change one word or one timing without starting over.

1. Preview. preview_video_frames checks the file, captures frames at the moments that matter and returns a sheet of them. Text that runs off the frame or collides with something comes back as a layout problem with the exact text. A visual review adds what a person would notice: text too small to read on a phone, a heading that leaves one word alone on its last line.
2. Fix and repeat. The agent changes what was flagged and previews again. Two or three rounds is typical.
3. Render. render_video makes the MP4, checks that its size and length match what was asked for, and saves it to Creative.

Images in the video, such as your logo or a product screenshot, come from your brand assets or files you've uploaded.

---

In a loop

Because the video is a file the agent writes, it fits a loop. Our own daily shipping loop now works like this every morning:

1. Read what shipped since yesterday and write the Discord, X and LinkedIn copy.
2. Make a 15 to 20 second 4:5 video of the same updates, using the same words.
3. Post to Discord, post to X with the video, post to LinkedIn with the video.
4. Send the links and the video to the team for a look.

If the video step fails, the posts still go out as text, and the run says why. In a loop the agent can only post to the accounts and pages you allowed when you set it up.

---

What it doesn't do (yet)

• No sound. Videos are silent. Add music or a voiceover in an editor if you need them.
• No generated footage. For people and scenes, use AI video (generate_video).
• 60 seconds at most per video.
• Your own images only. The video can use your brand assets and uploads, not arbitrary links from the web.

---

Try it

In Cogny chat: "Make a 15 second 4:5 video announcing [your update], in our brand colours with the logo at the end." The agent previews, fixes, renders, and shows you the video. Say "post it to X" and it asks you to approve the post first.