In development

This is a look at where Lumière is headed — not something you can buy today. We show it so you know our direction; we don't sell it until it ships.

Lumière · On the bench

Name it once.
The suite remembers.

People, animals, objects, buildings — the editor names what matters at ingest, and every later edit knows who and what is on screen. Opt-in, per client, deletable at any time.

The workflow we are building

Jog. Point. Name.

  1. 01

    Jog the raw footage

    Right at ingest, the editor scrubs through the new material — the same review pass you already do, nothing extra.

  2. 02

    Point at what matters

    A “Name this” button on the timeline: click or draw on the person, the dog, the product, the building in the frame.

  3. 03

    Type the name — once

    From then on the suite recognizes them across this edit, the next one, and every project for that client. “John”, “the dog”, “the office building” become words the story engine understands.

What that unlocks: transcripts and subtitles carry real names, the story engine arranges scenes around real people, dubbing keeps each person’s voice — and a written brief can simply say “the dog”.

The honesty rules

Quality gates, not guesses.

  • Unclear is never auto-judged

    A tiny, blurred or half-turned face goes to an “unidentified” review strip for the editor to assign by hand — the engine never labels what it cannot clearly see, and low-quality crops never train the memory.

  • Clear but uncertain asks first

    When the match is plausible but not strong, the suite asks “who is this?” instead of deciding. Only clear, confident matches are labeled automatically — and every automatic label stays revocable.

  • Voice supports, never overrides

    A recognized voice can suggest “is this John speaking?” — it corroborates the picture, it never overrules it.

  • Opt-in, per client, deletable

    The memory exists only where a client turns it on, holds only what the editor deliberately named, and can be deleted — entirely or name by name — at any time.

Where this stands

Honestly: in development.

The suite already tells speakers apart by voice within an edit and carries those labels into transcripts, subtitles and story briefs. What is in development is the durable, named memory described above — naming at ingest, recognition across projects, and the quality gates that keep it trustworthy. Until it ships and survives real client work, this page is a direction, not a product.