Today
Hundreds of hours of video sit on the drive: acceptance tests, machines at work on customer sites, assembly, trade fairs. Cutting one good clip from it often costs an editor a full day, so the footage stays where it is. LinkedIn and Instagram stay quiet for months, although the best shots were filmed long ago.
An example
Wertora Recycling has 340 hours of video on its NAS: acceptance tests, trade fairs and shredders at work on customer sites. For a series on the ZK 20 t/h shredder, the agent finds 26 strong scenes, cuts twelve vertical clips with German, English and Polish subtitles and flags three scenes in which someone stands at the machine without hearing protection. Marketing approves nine clips for the next eight weeks.
Your data, wherever it lives
Your footage is scattered today: raw recordings on the NAS and old hard disks, trade fair and acceptance films in project folders, finished clips in the media library, customer releases in the CRM. demi reads, understands and links it all scene by scene, with your rights. The AI cuts in minutes the clips an editor would need days for, and you decide what gets posted.
Where the data lives today
- Photos, video & audio
- NAS & hard disks
- Folders & drives
- CRM
- Social media
How demi solves it
Three layers in this order: connect the data first, then the rights, then the AI.
- 1 · ConnectData Layer
Data sources in this case: the media library, NAS, project folders, the CRM and your social media channels. demi replaces nothing and indexes every video scene by scene: which machine, which site, which customer, what can be seen and heard. Finished clips go into the media library and the content plan only after approval.
- 2 · PermitPermission Layer
The agent uses footage from customer sites only when the customer's release is recorded in the CRM. It flags scenes with recognisable people so someone checks consent. Only what marketing has approved gets posted, and every approval is recorded.
- 3 · ActIntelligence Layer
The agent describes each scene, rates picture, sound and message and proposes cuts in vertical, square and landscape format. It writes subtitles from the soundtrack in every language and takes technical details such as throughput or drive power from your data sheets. Scenes where someone stands at the machine without protective equipment are flagged, not used.
How it runs
- 1
demi reads the video archive on the NAS, drives and in the media library and indexes every scene.
- 2
You say what it should be about, for example: the shredder taking bulky waste, for Instagram and LinkedIn.
- 3
The agent proposes the strongest scenes and cuts clips in the right format, with subtitles and an end card.
- 4
It flags scenes without a customer release or with recognisable people for review.
- 5
Marketing checks, approves and schedules the clips in the content plan.
What changes
New clips from existing footage, regularly
Every channel in its own format
No post without customer release and consent
What does it bring you?
Check it with your own numbers
Every starting value is an assumption you can change. The calculation sits right below, step by step.
Your numbers
Result
+173 hfreed per year
Assumptions and calculation
demi finds the scenes and cuts subtitled clips; you check and post them.
3 h → 1.8 h per case
How we calculate
12 short videos × 3 h × 40% × 1.0 (manual) × 12 months = 173 h
Value per year, before demi costs
Freed per year+173 h
Assumption: labour cost of €49.50 per hour in German manufacturing in 2025 (Destatis), plus workplace, IT and overheads.
= Value per year, before demi costs
+€11,245
gained per year
173 h ÷ 1,760 h per role = 0.1 full-time roles
An estimate, not a promise. The starting values are assumptions for an industrial company with 150 to 500 people.
Combine with more cases and get the packageQuestions about this use case
How does demi find the good scenes in hundreds of hours of footage?
The agent indexes every video scene by scene: what can be seen, which machine is running, what is said, whether the picture is steady and sharp. For your topic it proposes matching scenes, each with file and timestamp. You decide which ones go into the clip.
What about people and customer plants in the picture?
The agent flags scenes with recognisable people so you can check consent. It uses footage from customer sites only when the customer's release is recorded in the CRM; otherwise those scenes stay locked. So nothing goes out that shouldn't.
Does every channel get its own format?
Yes. The same scene becomes vertical for reels and stories, square for the feed and landscape for YouTube, each in the right length, with subtitles and an end card from your template. The agent writes the captions per channel and language.
Does our raw footage have to move?
No. demi reads the videos where they are: on the NAS, on drives, in project folders or in the media library. After approval, finished clips go into the media library with a link to the original scenes.
