How to work through it
Write the audience, communication objective, offer, approved message, mandatory legal copy, prohibited claims, brand codes, delivery placements, and measurement plan. Attach current identity assets and document the rights for products, people, voices, music, footage, and stylistic references. Give each multimodal source a declared role so a mood clip cannot quietly override an approved logo, packaging detail, spokesperson identity, or product fact.
Break the concept into a few beats: recognizable opening, product or idea reveal, supported proof, emotional turn, and accurate action. Specify camera, performance, timing, dialogue, music, ambience, and silence as connected decisions. H3's official materials document stereo audio-video output, but the final mix, pronunciation, captions, territory rights, and platform loudness remain production responsibilities that need human review.
Review logo geometry, typography, colors, packaging, product behavior, spokesperson likeness, voice, cultural context, subtitles, and every implied comparison. Separate expressive imagery from factual demonstration. If one beat fails, revise the smallest input relationship or direction while preserving approved elements. Do not select a visually striking output that changes the offer, invents a product behavior, imitates an unauthorized person, or undermines required disclosure.
Export variants only after checking crop, duration, first-frame clarity, muted viewing, captions, audio, accessibility, landing-page parity, regional requirements, and platform synthetic-media policies. Save the prompt, source manifest, H3 path, output, edit history, approvals, disclosure decision, and expiry date. Measure the released creative with the campaign's own experiment design rather than attributing an outcome to the model without evidence.
