Turn Image Prompt Optimizer into a reviewable shot contract.

Write the invariant shot intent first

For Image Prompt Optimizer, state the subject, environment, action objective, emotional beat, and end condition in plain production language. These invariants keep general video prompts variants comparable instead of producing unrelated ideas. For this section 1, save a prompt brief; have the producer review source rights; record the failed case as well as the accepted one; and do not advance it beyond a editorial approval until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.

Turn the topic into a repeatable method

Move from definition to a small practice sequence, then review the result against explicit craft, rights, and delivery checks. Keep vendor-specific behavior outside the method unless a source verifies it. For this section 2, save a evidence ledger; have the brand reviewer review revision control; record the failed case as well as the accepted one; and do not advance it beyond a editorial approval until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.

Layer camera, timing, continuity, and sound

Add framing and camera movement only when they support the action. Mark temporal beats, preserve identity and environment anchors, describe sound as intent rather than a guaranteed feature, and name artifacts or changes to avoid. For this section 3, save a continuity sheet; have the motion designer review motion coherence; record the failed case as well as the accepted one; and do not advance it beyond a editorial approval until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.

• Subject and action — record versioned handoff, post supervisor, visual hierarchy, and dated decision; keep optimizer evidence separate from image assumptions.
• Framing and camera relationship — record asset ledger, creative producer, camera intent, and delivery package; keep optimizer evidence separate from image assumptions.
• Beat order and end state — record decision memo, prompt designer, claim support, and delivery package; keep optimizer evidence separate from image assumptions.
• Continuity, sound intent, and exclusions — record acceptance matrix, prompt designer, input fidelity, and dated decision; keep optimizer evidence separate from image assumptions.

Create controlled prompt variants

Change one meaningful dimension per version, such as shot size, movement, pacing, or performance intensity. Label the hypothesis for each variant so review can identify which instruction caused the difference. For this section 4, save a annotated source board; have the visual lead review licensing; record the failed case as well as the accepted one; and do not advance it beyond a editorial approval until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.

Review the output without inventing model controls

Check whether the requested action reads clearly, the camera supports the beat, references remain coherent, and the ending is usable. A prompt template does not prove that any named model, duration, audio mode, or parameter is available. For this section 5, save a test worksheet; have the fact checker review input fidelity; record the failed case as well as the accepted one; and do not advance it beyond a editorial approval until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.

Keep the evidence ledger attached to the decision

Partial evidence was supplied, but it does not establish product support or a complete capability, customer, or performance claim. Record the source, verification date, claim scope, unresolved gap, and the decision that the evidence can support. Search demand must never be reused as capability proof. For this section 6, save a revision log; have the prompt designer review editability; record the failed case as well as the accepted one; and do not advance it beyond a editorial approval until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.

Build a specific test brief for image prompt optimizer

Start with an authorized still image and its untouched original. Define one observable change, protected details, a stopping rule, and the named reviewer. The intended output is a reusable prompt contract with controlled variants. Test one variable per version, preserve the source and settings, and compare results at the actual delivery size instead of choosing from an unrecorded impression. For this topic test, save a camera plan; have the brand reviewer review editability; record the failed case as well as the accepted one; and do not advance it beyond a versioned review until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.

• Primary query: image prompt optimizer; test record: prompt brief, visual lead, licensing, and dated decision; keep optimizer evidence separate from image assumptions.
• Editorial owner: keyword-expansion:0554; decision record: delivery checklist, technical reviewer, temporal stability, and workflow decision; keep optimizer evidence separate from image assumptions.
• Source scope: a user-provided competitor-gap export dated 2026-08-05 supports topic prioritization only; source review: decision memo, visual lead, temporal stability, and approved master; keep optimizer evidence separate from image assumptions.

Separate topic fit from product proof

A dated, user-provided competitor-gap export supports only the decision to cover “image prompt optimizer.” It does not prove audience demand, SEELE capability, third-party behavior, commercial value, or a likely outcome. Verify product-specific statements against current first-party documentation and a recorded representative test. For this source review, save a annotated source board; have the post supervisor review source rights; record the failed case as well as the accepted one; and do not advance it beyond a rights-cleared draft until the named limitation is resolved. Keep optimizer evidence separate from image assumptions.