Spectre: an AI platform for reality TV

A reality production generates thousands of hours of video and audio. Somewhere in there are the moments that make a season: a confession, a break-up, one conversation that changes everything. They happen once, and someone has to find them again. Spectre helps with that. Your Tech Club built the platform together with Signal.Stream, which makes content for television and streaming.

What needed to get better?

Searching takes days. Story teams watch hours of dailies to find the scenes editors build around. General-purpose AI models barely help. They are trained on books, code and the internet, not on television, and they return words rather than a moment. They don't know what a confessional sounds like, they lose the thread when six people talk over each other, and they don't know the rules of a format.

On top of that, unreleased footage is sensitive. You don't put footage from a season that has yet to air into a public AI tool.

Two sides that needed each other

Production know-how without technology would have stayed an idea. Technology without production context would have solved the wrong problem. Signal.Stream brought years of experience in television production, knowledge of editorial workflows and access to real productions. Your Tech Club brought the AI engineering, the design and architecture of the platform, and the engineering to process large volumes of material.

That led to the most important decision: not to adapt a general-purpose model, but to build a system for one industry. Spectre learns the cast, the format and the show bible of a production, and looks for what matters to that particular story.

How we approached it

Spectre was shaped from the start together with editors, story producers, showrunners and production teams. Their feedback decided not only what the technology could do, but also what it had to do. Every feature is designed around a problem they run into every day, with real productions as the test.

What did we build?

A platform that turns raw video and audio into a story a team can work with. It runs on a proprietary model built for unscripted television.

Scenes, emotions and conversations

Spectre detects and organises scenes across every camera angle and shooting day, flags emotional turning points, conflict and tension, and maps who said what to whom. Alliances and conflicts become visible, without spreadsheets or relying on memory.

Built for multiple speakers

Where a general-purpose model gives up when six people talk at once, Spectre separates the voices and follows speakers across days and cameras. Every moment comes with the exact second it happened, so an editor lands on the right spot in the timeline straight away.

An assistant for the story team

Daily summaries of what happened and what matters, and an AI Spotter trained on the production's show bible and format. Spectre understands Dutch and the main European languages, for productions that aren't made in English only.

Footage stays with the production

Every production runs in its own isolated environment. Material is not used to train models for other productions, it is encrypted in transit, at rest and during processing, and every access and action is logged. A production company chooses the region where its data is stored, and Spectre can run in a private cloud, on-premise or on infrastructure the production manages itself.

The website

Spectre's website was also designed and built by Your Tech Club, in Next.js. Production teams request access there under NDA; new places open in rounds.

Where does it stand now?

Scene detection, conversation intelligence, emotion and tension detection, daily reports and the AI Spotter are live. The next step is on the roadmap: rough cuts ready for the edit, direct export to Premiere Pro and DaVinci Resolve, and timelines organised by storyline.

Spectre doesn't replace an editor's instinct. It takes the searching away, so a team can spend its time on the story.