Guides

Build prompts around subject, environment, composition, light, material, lens, and finish instead of adjective piles.

Use clear product identity, controlled sets, realistic materials, intentional shadows, and reference-led iteration.

Lock the visual world first, then vary composition and format while preserving lighting, palette, material, and lens language.

Choose an approved source image, describe only believable motion, and separate subject action from camera movement.

Use varied clean references, define a credible professional context, and protect natural skin and facial structure.

Choose format from the publishing surface first, then compose for crop safety and focal hierarchy.

Match model strengths to photography, text, references, editing, speed, and budget instead of using one default.

Compare native audio, reference support, duration, motion quality, adherence, and cost before rendering.

Decide whether a reference controls identity, composition, material, color, or style, and state what must remain unchanged.

Track model terms, model source, prompts, references, approvals, and delivery context for every commercial output.

Move from brief to visual territory, controlled variants, comparison, selection, refinement, and organized delivery.

Store the prompt, model, settings, reference, cost, campaign, status, and final decision with the media itself.

Use the same brief, format, references, and decision criteria, then compare full outputs rather than cherry-picked thumbnails.

Choose the model deliberately, batch only when variation has decision value, and retry failed jobs without duplicating successful work.