20:56 to 21:00
Group game at the fire pit. Mia jokes about yesterday's vote. Ana goes quiet.
Live on set, from the first take
AI production assistant for unscripted TV.
Faster stories, ready-made clips, and the drivers of what works. IMR follows the shoot as it happens and logs the story from transcripts and scene context, so your teams review instead of logging.
Feeds · 6 of 6
CAM 1
CONF
CAM 3 · SPEAKING
CAM 4
CAM 5
CAM 6 · EMPTY
Transcript · CAM 3Auto · EN
- 21:03:58 Mia I'm just saying, nobody told me the rules changed.
- 21:04:09 Luka They didn't change. You weren't here yesterday.
- 21:04:21 Ana Luka. Can I borrow you for a second?
Story logEvery 2 to 5 min
20:52 to 20:56
Kitchen. Luka and Jan plan a surprise for the weekend.
Today · 14 h recorded · 6 feeds Highlights marked by IMR
Illustrative interface. Cast, dialogue and data are invented.
01·The problem
Every show is logged and clipped by hand. Audience data never reaches the storyline.
01 · Live production
Logged in spreadsheets
Story editors watch every camera, transcribe what happens and type it into a log. Pre-cuts, rough cuts and final cuts are all decided by hand.
02 · Social media
Watched twice
The social team watches the whole show again to find the moments, then cuts, captions and reframes each one for vertical video.
03 · Audience
Read too late
Ratings, streaming and social reactions arrive after broadcast with few indicators. None of it shapes the episodes still being cut.
TC inTC outWhoWhat
20:52:1020:55:48Luka, Jankitchen, planning something?? check cam 2
20:56:0221:00:30allfire pit game, mia joke re vote (good?)
21:04:Ana, Lukahallway, ana takes luka aside. sth about the vote?? check cam 3 + 4
02·How it works
One loop. It speeds up production and gets smarter with every episode.
Four layers, each built on the one before. What IMR learns from each episode, and from every clip it posts, feeds the next.
Index
Turns hours of footage into a searchable story log, while the shoot is still running.
- Transcripts split into scenes and meaningful units
- Live summaries every 2 to 5 minutes
- Plot threads followed across episodes
- Search by dialogue, cast or description
- Pre-cuts built from the summaries Next
Create
Turns raw footage into vertical short-form edits the social team approves instead of cutting.
- The strongest scenes and sequences, found for you
- Reframed to 9:16, following who matters
- Highlights by cast member
- Edits built on story patterns IMR finds
Setup → punchlineClaim → contradictionBefore → afterLine → reactionsRunning gags
Analyze
Gives a full, data-based picture of every episode, scene by scene.
- Virality and tune-out risk per scene
- Screen time and scenes per cast member, cut rate
- Picture checks: brightness, black and frozen frames
- Ratings, streaming and social data side by side
Scores are not yet validated against ratings.
Every posted clip reports back
When a clip IMR cut goes live, its reach, likes and comments come back to the scene it was cut from. IMR already knows who is in it, which thread it belongs to and which pattern it used, so the numbers explain something.
- Reach, likes, shares and comments for every clip
- Comments read against the scene’s transcript and threads
- Results compared by pattern, cast member and thread
Turning these results into the next edits is the Drive phase.
Drive
Will find what drives engagement and feed it back into the next episodes and social edits.
- Tests story factors against ratings, streaming and social data
- Recommends storylines to push next
- What audiences responded to shapes the next pre-cuts and clips
Not yet proven. It needs validation against real audience outcomes, and that is our next phase.
All interfaces on this page are illustrative. Cast, comments and figures are invented.
03·Why it works
IMR reads context.
Transcripts, scene context and many per-scene indicators combine into one story model. So IMR understands who is in conflict, what is building and why a moment matters.
Transcript
Ana · "We need to talk. Not here."
Scene context
Hallway, night · Ana and Luka in frame · minutes after Mia’s challenge
Per-scene indicators
- Who is in conflict
- Ana and Luka, over the vote two episodes ago.
- What is building
- Their third private talk this week, and the first since Mia arrived.
- Why it matters
- It pays off Ana’s confessional from episode 22. Worth a pre-cut and a Claim → contradiction clip.
04·Where we are
What works today, and what comes next.
Live story editing
Summaries every 2 to 5 minutes from the show’s live audio.
- Used by story editors to pick storylines
- Next: editors ask for the storylines they want
- Next: pre-cuts built from the summaries
EvidenceLive production test
Social clips
Clip candidates reframed to 9:16, following who matters in the scene.
- Highlights by cast member, ready to post
- Also works on footage shot only for social
- Approve instead of cut
EvidenceIn use at a major broadcaster
Full season
A full season of a reality show is being processed.
- Several languages on one platform
- Per-scene, per-cast and per-episode data collected
- Ratings data ingest built
EvidenceSeason in progress
Next
Not yet proven: that audience data shows what drives engagement. That is the next phase.
Product roadmap
From two layers in use to the full production cycle.
Today
- Live story editing, tested on a live production
- Social clips, in use at a major broadcaster
- A full season being processed
05·Team
Technical depth and business experience, from day one.
Tech · Co-founder
Grgur Damiani
- Builds AI products for television
- Built this platform in about five months, part-time
- Background in technology consulting and implementation
- Data Science and AI, Maastricht University
Business · Co-founder
Fran Stilinović
- Runs value creation initiatives in private equity portfolio companies
- Background in digital strategy and consulting
- Winner of Future Entrepreneur 2019 (Ekonomska klinika)
- BSc Economics, Management and Computer Science, Università Bocconi (Bocconi International Award)
Want to build AI for television? We are a small team and will grow it carefully. Tell us what you would build.
Write to us about joining