0:20 · Your First Generation

Every generation in Loft, everywhere in the app, is the same four beats: pick a model → set up inputs and prompt → Generate → watch the queue. Learn it once here and the Assistant, the Image Builder, and the Flows canvas are all just different front doors to it.

Open your asset's Generate tab. It opens on a model gallery headed "Select a model to get started."

There's no separate "what do I want to make?" switch — the model carries its modality, so choosing a model is choosing whether you get an image, a video, audio, or 3D. Narrow the list with:

• The search box, to filter by name or description
• The type pills — Text to Image, Image to Video, Text to Audio, Upscale, and so on
• Recently Used, which floats your habitual models to the top as you build a history

For this lesson pick a Text to Image model. Click its card; the view splits into a setup form on the left and a live queue on the right. ← Back in the header returns you to the gallery.

A vague prompt gets you a competent, generic image. Name four things and you'll get something you can actually direct:

• Subject — what's in the frame
• Framing — wide establishing, close-up, three-quarter, overhead
• Light — golden hour, hard noon sun, single practical in a dark room
• Treatment — the medium or look: 35mm film, matte painting, product photography

A workable first prompt: "wide establishing shot of a coastal city at dusk, low golden light raking across the buildings, 35mm film, slight haze."

Above the prompt box is a small toolbar — Primers (reusable prompt blocks), Save current prompt as primer, and Browse previous prompts. Ignore them this pass; you'll use them next lesson.

Parameters are generated from the model's own definition, so you only ever see controls that model actually supports. Seed, steps, and guidance are tucked behind an Advanced disclosure. Defaults are sensible — don't open it.

One control does matter: Number of Images defaults to 1. Keep it there. While you're still finding the prompt, four variations of the wrong idea cost four times as much and teach you nothing. Raise the count later, once the setup is right.

A live cost estimate sits next to the prompt toolbar — in credits or dollars depending on your billing mode. Glance at it. Then press Generate.

The job appears in the queue on the right and moves through Pending → Running → Complete, refreshing every few seconds. Nothing is lost if you navigate away — the queue is a real record, not a progress bar. Watch it from My Queue in the sidebar too.

When the job completes, thumbnails appear on the queue card. Click one for the full-screen preview. Then expand the card — this is the part worth pausing on. You get:

• Copy prompt and Regenerate — your iteration handles for the next lesson
• Open in flow — the branch icon jumps to the lineage canvas for the outputs
• The inputs, grouped by the role each one played
• A details table: model, size, seed, steps, and the rest

Loft kept all of that. The output also landed in your asset's Files as a new asset item — the link on the card jumps you there. Nothing you generate is ever an orphan file with a hash for a name.

• No Generate tab — add it with + Add tool on the asset. (The tab needs a server-side asset; a purely local, unpublished asset won't have it.)
• Generate button disabled — your balance won't cover the estimate. The panel shows credits needed vs. available, with a Purchase credits link.
• Job shows Error — expand the card to see why. A different model, or a shorter prompt, is usually the fastest recovery.
• Nothing in the queue — check you're looking at the right project; the queue is project-scoped.

• A queue card reads Complete and shows a thumbnail.
• The same image appears in your asset's Files tool with status New.
• You can expand the queue card and see the model and seed that produced it.