Nobody wrote it down. Granola did.
Granola Briefs reads everything that happened between meetings — the recordings, the threads, the emails — and writes you a short brief before the next call starts. The problem it solves isn't missing information. It's that the information exists in five places and you have four minutes. This film had to make relief feel earned, not announced.
Self-initiated concept — Granola Briefs
• Concept & script • Visual asset creation •
• AI cinematography direction •
• Motion Design & Animation •
• Sound Design •
About the project
THE PRODUCT
Granola is an AI notetaker that sits quietly in your meetings and writes the notes for you. Briefs is the part that runs before the meeting: it reads the recordings, the Slack threads and the emails from the last few weeks and hands you one short document — what was decided, what's still open, what you owe someone. Everything you agreed with the client, in one place, before you walk in.
THE PROBLEM
Every AI notetaking product says the same three things: it records, it summarises, it saves you time. Briefs actually prepares you for the meeting. But you can't film a document being ready. So I focused on the person instead of the product. I wanted to show how someone actually feels before a call they're not ready for — and how they feel after. That change is the real thing the product does. A summary is just a summary. Walking in calm is something people recognise.
That's why the camera stays on his face longer than on the screen, and why the film ends on him breathing out instead of on the brief.
MY PART
Everything: the insight, the script, the shot list, the asset design, the AI direction, the animation, the sound design, the edit.
AI generated the actor, the light and the room from my prompts. Getting the look right took many iterations and a lot of reference gathering, so the generated frames matched what I had in my head. The UI was animated by hand in After Effects.
Design Decisions
The screen shows the three sources
— a recording, a Slack thread, an email. So the voice-over doesn't list them. It says the one thing the image can't: "Nobody wrote it down." The product answers in two words: "Granola did."
For the first eight seconds there's no music, just a low bass building tension. The music only enters when the brief appears — warm and hopeful — so the shift in sound does the same thing as the shift in how he feels.
I used sound instead of graphics to signal the meeting coming up. A notification tone reads faster than a banner on screen, and it doesn't ask the viewer to stop looking at his face.
The recording is dated Jul 14, the Slack thread Jul 28, the email Aug 3. Each one feeds one section of the final document, so nothing in that brief looks invented.
Where AI ended and the work began
AI: the actor, the light and the room
— a casting session and a location I couldn't have afforded. One approved reference frame became the source for every shot, which is the only way the same man appears throughout without drifting. It took twenty-eight generations to get there; the rejections all failed the same way — skin without pores, light coming from everywhere at once,
a screen glowing blue on a face. Voice-over and sound effects generated in ElevenLabs.
By hand: everything the product does. The interface, the asset cards, the brief writing itself line by line, the timing, the compositing, the grain and colour matching, the mix and sound design. Built in Photoshop and After Effects, edited and mixed in Premiere Pro.
Explore more motion projects ➔
Some people have a safety net.
She has Manus.
Manus is an agent you hand a goal to. It searches, reads, compares and writes the result while you do something else. The camera stays on her longer than on the product. You have to feel the problem before the solution means anything.
Some people have a safety net.
She has Manus.
Manus is an agent you hand a goal to. It searches, reads, compares and writes the result while you do something else. The camera stays on her longer than on the product. You have to feel the problem before the solution means anything.
Self-initiated concept — Manus
• Concept & script • Visual asset creation • AI cinematography direction •
• Motion Design & Animation • Sound Design •
Self-initiated concept — Manus
• Concept & script • Visual asset creation •
• AI cinematography direction •
• Motion Design & Animation •
• Sound Design •
About the Project
THE PRODUCT
Manus is an AI agent. You give it a goal and it handles the steps on its own - searching, opening sources, pulling out what matters, comparing it and writing the findings into a finished document. You write one sentence. It does the job in the background and hands you the result.
THE PROBLEM
An agent is difficult to film. The work happens somewhere nobody can watch it. Most films in this category solve that by putting the machine on screen: tabs opening, tasks ticking off, progress bars filling. I watched a lot of them before I started and none of them made me want the product. Software thinking is boring to look at.
MY PART
Everything: the insight, the script, the shot list, the UI design, the AI direction, the animation, the sound design, the edit. AI generated the woman, the beach and the light from my prompts. Every interface in the film was designed and animated by hand in After Effects.
Design Decisions
She says one line out loud, and the other one we watch typing itself into the Manus search panel. "Oh god, I totally forgot" is spoken, because that makes the panic she feels hit harder. The command is shown as text being typed into Manus, because the viewer has to understand that this is the moment Manus starts gathering the data and the whole magic begins.
In the first few seconds I focused on her. That's the only way the viewer relates to what she's going through. And when someone relates to the character and her problem, they're much more likely to notice the product we're advertising and want it - in this case Manus, which gets you out of trouble.
I kept the research part short. The point wasn't for the viewer to read everything, just to see that a lot of work is happening at once. If I'd shown it properly and made it all readable, it would have taken too long and people would have scrolled past.
Where AI ended and the work began
AI: AI: the woman, the beach, the lipsync. Consistent frames with a consistent character, built on a prompt I refined over many iterations — I wanted her to look as natural as possible. Voice-over and sound effects generated in ElevenLabs. Frames and AI-generated video in Higgsfield.
By hand: everything the product does. The message layer, the Manus input field, the source cards, the extraction sequence, the report, the send and sent confirmations, the compositing, the timing and the mix. Designed in Photoshop and Illustrator, animated in After Effects, edited and mixed in Premiere Pro.
The competitor names, feature sets and report contents are fictional assets I designed for the film.
Unofficial concept - not affiliated with or commissioned by Manus.
Explore more motion projects ➔
Explore more motion projects ➔