Core Concepts

Loft is built around intent: you describe what you want and the Assistant does the work — routing to the right model, running jobs, and recording the result. Instead of wiring a pipeline by hand, you converse with it, and the durable artifacts (assets, generations, lineage) accumulate as you go. Learning the handful of nouns below makes everything else click.

• Project — a body of work. It runs in Virtual (cloud) or Local (your folder) mode.
• Library (category) — a group within a project: Characters, Environments, Shots, Editorial, etc. These are the entries under LIBRARIES in the sidebar.
• Asset — a single thing in a library: a character, a shot, a cut, a daily. It has metadata, tasks, notes, and a set of tools (tabs).
• Asset item — an individual file inside an asset (an image, a video, an annotation). Assets hold many items.
• Take / version — successive results for the same asset; the review status marks which ones matter.

Every file carries a status that drives the rest of the pipeline:

• New — not yet reviewed
• Selected — a candidate you've picked
• Published — the finalized, deliverable take (Editorial conforms to the latest Published take of each shot)
• Dropped — rejected
• Reference — an internal input for generation, not a deliverable

• Assistant — the conversational agent (header button). Finds things, answers questions, generates, and takes action across the project.
• Queue — every generation job lands here; it's where you watch progress and inspect parameters.
• Playbooks — standing guidance the Assistant reads every turn (project / asset / user / model), so you teach it your conventions once.
• Lineage — each derived item records what it came from, so a generation's history is never lost.