An AI wedding film should be planned around a relationship and an ending, not a list of romantic camera moves. Approve the story first, keep separate visual references for both partners, and generate manageable shots that can be reviewed individually.
A stylized film made from portraits is not footage of an actual ceremony. Make that distinction clear when sharing it, particularly if the location or actions are invented.
1. Set the promise and delivery
This illustrative project is a 40-second horizontal creative film for the couple: “Two people cross a quiet courtyard and leave together.” Its setting combines a stone courtyard with Western-inspired wardrobe accents. It is fantasy art direction, not a claim of historical accuracy.
Use authorized photographs and agree on clothing, permitted changes and how the result will be shared. The plan has not been rendered or evaluated as a completed film.
2. Approve a treatment with a visible ending
The partners arrive from opposite sides, notice each other, meet, join hands and walk toward an open gate. The final image is the pair moving together, not a random landscape.
Keep vows and voice-over separate. Do not invent a quotation and present it as something the couple actually said. Any use of a person's voice needs appropriate permission and supported tooling.
3. Lock the continuity bible
Use separate identity references for partner A and partner B. Lock each hairstyle, outfit, accessory and relative height. Define the bouquet's size, which hand holds it, where the sun is, and which direction the pair will move.
A dark suit, ivory dress and small bouquet provide clearer visual anchors than changing outfits in every shot. If the creative brief includes boots or a hat, specify who wears them and when.
4. Give every shot one job
| Time | Story purpose | Action and camera | Stable endpoint |
|---|---|---|---|
| 0–8 s | Establish separation | Wide courtyard; restrained lateral move | Each partner holds a clear position |
| 8–16 s | Recognition | Medium view of partner A turning toward B | Gaze settles toward the correct side |
| 16–24 s | Approach | Wider two-person view; short controlled walk | Both stop with coherent foot contact |
| 24–32 s | Connection | Medium close view; one hand joins another | Hands and bouquet remain distinct |
| 32–40 s | Resolution | Rear three-quarter view; walk toward the gate | Pair settles into a usable closing frame |
Eight seconds is a planning duration here. Check actual provider limits before turning this into requests.
5. Map references and handoffs
Re-anchor from approved couple and courtyard keyframes at scene boundaries. Use accepted endpoints for short within-scene continuity only; return to an approved reference whenever drift appears or the planned scene changes. Do not use a frame with a distorted hand merely because it is the last frame.
6. Prepare the next unresolved shot
For shot one, specify the approved courtyard layout, the two approved identities and outfits, their starting positions, one restrained camera move and a stable end composition. Confirm reference-role support before compiling a provider request. Later shots remain plans until the relevant review is complete.
The motion-focused structure in Runway's guide is a useful reference, but provider-specific features should not be assumed transferable.
7. Apply a couple-specific quality gate
Check each face individually, then relative height, eye-lines, hands, bouquet, clothing and movement direction. A shot fails if two recognizable people become similar generic faces, even when the setting is attractive.
Review the emotional sequence too: does recognition precede approach, and does the ending resolve the action? Assemble sound and final color only after the picture works.
8. Request the next approval
Approve the treatment before rendering, then accept or revise each shot. Imagild's film workflow is intended for this staged approach; a single rejected clip should not require abandoning every accepted shot.
Will matching clothing guarantee identity? No. It is one continuity aid, not facial verification.
Should I generate a complex kiss or ring exchange first? Start with simpler interaction. Close hand and face contact requires particularly careful review.
Filled decision record: storyboard purpose
This is a constructed review exercise, not a reported generation, software test or audience study. Its purpose is to make the acceptance decision reproducible for the next reviewer.
| Review item | Hypothetical evidence or requirement | Decision and next action |
|---|---|---|
| Failure observed in the candidate | The meeting has an attractive courtyard but no understandable resolution. | Hold the affected candidate; name this failure before changing anything else. |
| Protected result | End on the pair leaving together. | Compare against the approved source and intended sequence, not the attractiveness of an isolated frame. |
| Bounded next action | Remove the unrelated landscape ending and review the action sequence without music. | Make this change first, then inspect the same acceptance criterion again. |
Keep the actual source version and the rejected candidate distinguishable. A planning correction does not prove a later output follows it, and a successful individual frame does not approve the complete clip or sequence. Record which requirement is still unresolved rather than silently labeling the whole deliverable final.
For the next handoff, include the approved requirement, the evidence that supports it and this specific unresolved action. This record belongs to the editorial workflow; it is not a claim that Imagild enforces these constraints automatically.
Documentation and production boundary
LTX Studio: Script to Video is a vendor-specific workflow reference. The decision record above is our original manual review method. Use Imagild's current film planning workspace for supported planning steps and inspect the current interface before choosing an input or generation mode. Do not infer an available control from another vendor's documentation.