Lesson Outcome
A strong prompt is not yet a strong frame. It can sound confident, promise ideal lighting, and preserve every important word while the actual image grows a hand out of a cup. You do not accept an elegant instruction. You accept a file that passed inspection.
In this lesson, you will prepare an image prompt, run one small test, select the actual file, and decide whether it is usable for the next step. If a project needs several anchors, you can collect verified frames into a Frame Pack. To begin, one scene and four lines are enough: the moment needed, what must be visible, what must not change, and what the frame will be used for next.
Full timing, a spatial map, a shot plan, and a reference set provide more control, but one simple frame should not become paperwork for paperwork’s sake.
First Choose One of Three Tasks
A new frame is for when no image exists and you need to create the scene, subject, composition, and lighting. A frame from existing authorities is for when character, product, or location references exist and you need one specific view without mixing their roles. A local edit is for an almost-finished image: change one thing and preserve everything else.
Choose one task before writing the prompt. A new viewpoint needs a fresh description of the visible part of the scene. A local edit needs to hold carefully to the parent frame instead of reinventing the room.
World, Frame, State, and Task
The world is what exists and where it is. The frame is the part of that world the camera shows now. State is what has already happened to characters and objects. The task is what you want to create or change right now.
A phone may lie on the table without appearing in a close-up of a face. Do not force the whole world into every prompt or announce that a hidden prop disappeared. Describe the visible part, the needed state, and the next concrete task.
When you use several references, label each role in one line: portrait for face; jacket photo for wardrobe; room frame for geometry; advertising example for lighting only. If sources conflict, choose the primary one before generating.
Build the Image Prompt
Build the description in the order a viewer reads the frame: who or what is in it; state and pose; visible geometry and composition; shot size and angle; lighting, materials, and environment; what to preserve; and one or two specific risks. Keep format, resolution, and similar settings in the interface when it provides dedicated fields.
Subject and visible state:
Scene geometry and composition:
Shot size and angle:
Lighting, materials, and environment:
What to preserve:
Specific risks to avoid:Full Example
This is a prompt to copy into an image generator. It describes one still moment, not a hidden video.
A naturalistic live-action medium shot of a job candidate seated at the near side of a narrow rectangular table, facing the recruiter across from him. The candidate has just noticed a single white sheet lying closer to the recruiter. Preserve the approved character identity, charcoal jacket, room geometry, opposite chair positions, and the door behind the recruiter's right shoulder. The candidate's right hand rests near the table edge without touching the paper. Soft side light enters from screen-left, with believable skin texture, restrained contrast, real fabric and wood materials, crisp stop-frame clarity, and no motion blur. Keep the table axis and eyeline direction stable. No extra papers, no rearranged furniture, no synthetic CGI finish.Local Edit: One Operation
If only the jacket must change, do not rewrite the whole scene. Name the change, everything that must be preserved, the allowed edit area, and a visible sign of success.
What to change:
What to preserve:
What the edit may affect:
What success looks like:Change only the candidate's jacket from navy blue to matte charcoal black. Preserve the face, hairstyle, pose, hands, paper position, table geometry, camera, framing, lighting, background, and all other visible details exactly as in the current image.A Frame for Future Motion
A beautiful photograph is not always a strong starting anchor. Future motion benefits from clear object boundaries, a readable pose and starting state, room for action, visible hands and contact points, no accidental blur, and stable face, product, text, and geometry.
If a character must lift a cup, the cup is still on the table in the starting frame, the hand has not already grabbed it, and the path remains clear. The next action then has somewhere to go rather than somewhere to escape from.
Real Test and Evidence
Open an available image tool and confirm it accepts the inputs you need. Run one small test, save the actual file under a clear name, view it at full size, and decide: accept or retry, with one reason.
If you have not selected a tool, do not search for “the best forever.” On official pages, find no more than three options that accept the required input and export the required file; remove unsuitable options and run the same small test on those that remain. That is enough for a first choice.
Moment and frame role:
Tool, model, or version, if shown:
Selected route and meaningful settings:
Materials used and their roles:
Prompt text:
Actual selected file:
What visual inspection confirmed:
What remains unproven:
Decision: accept / retry:
One reason for the decision:If a project needs several frames, repeat the cycle only for genuinely necessary start, key, or end states. One project may finish with one accepted file; another may need several. A frame pack is an extension, not a required ritual.
Check
If the prompt is exquisite but a hand grows from the cup, the prompt does not pass. The frame does.
Takeaway: you have an actual selected and verified frame, or an honest decision to retry one specific part.