# Make the flamingo lagoon ad with Genjutsu

The workflow, reusable prompts and soundtrack-restoration steps behind my pink flamingo Short.

[HTML](https://lukejbyrne.com/library/guides/vbe3lluhqxs/)
Content language: en
Author: Luke Byrne
Video published: 2026-09-12
Guide published: 2026-10-31

[Watch the video](https://www.youtube.com/watch?v=vbE3LluHQXs)
Video language: English

- [Download the English PDF guide](https://lukejbyrne.com/library/downloads/higgs-flamingo-ad-guide.pdf)

## Set up the shot

Start with a short reference video and a clear portrait. Use footage, likenesses and audio you have permission to reuse, and keep the original soundtrack. The saved September production used Genjutsu through MCP at 720p, with a 30-second input.

1. On Higgsfield, open Video, then Create Video, and choose Genjutsu. Its current official help page accepts 4-30 second reference videos.
2. Upload the video and portrait. On the website, choose Motion transfer for a new scene, or Objects swap for a targeted replacement.
3. Use the first prompt below to create the lagoon version. Inspect the face, hands, phone, water contact and coastline across the entire camera pullback.
4. Save the accepted lagoon video. Use it as the new input for Objects swap, then apply the flamingo prompt. Check the quoted cost before each generation.

## Prompt 1: face and location

Edit the reference video, keeping its continuous camera pullback, shot timing, framing, body actions and final fade. Replace the person with the adult in my portrait. Match the face naturally to the seated pose, clothing and light. Preserve the phone, hand gestures, head turns and floating motion. Rebuild the setting as a secluded volcanic lagoon: calm teal water, warm late-afternoon sunlight, black sand, basalt rocks and mossy cliffs. Reveal one small cedar cabin as the camera pulls back. Keep the geography consistent; add no cuts, crowds, boats, text or logos. Keep the donut float for this first pass.

## Prompt 2: replace only the float

Make one targeted change to the accepted lagoon video: replace the orange sprinkle donut float with a pink inflatable flamingo. Give it a ring-shaped body, a curved neck, a small head with a black-tipped beak, subtle wings and glossy vinyl. The ring must support the person in the same seated position and follow the existing bobbing and trajectory. Place the neck to the side, clear of the face and phone. Remove the donut sprinkles. Preserve the person, clothes, hands, phone, lagoon, cliffs, cabin, lighting, pullback, timing and final fade. Do not redesign the background or add cuts, people, text or logos.

Use the version you already like as the revision input. This gives the next generation a concrete lagoon, camera move and performance to preserve.

## Restore the soundtrack

The original workflow restored the source audio after generation. A visual prompt asking for the same music does not prove that the sound survived. With FFmpeg installed, put the generated picture in generated.mp4 and the permitted original video/audio in source.mp4, then run:

ffmpeg -i generated.mp4 -i source.mp4 -map 0:v:0 -map 1:a:0 -c copy -movflags +faststart final.mp4

This copies picture from the first file and audio from the second. It does not fix a duration mismatch. Compare durations first, then watch and listen from beginning to end; align the soundtrack in an editor if needed.

## What to check before exporting

Keep the face recognisable, the hands and phone coherent, and the float in believable contact with the water. The flamingo neck should stay clear of the face and phone. Watch the coastline and cabin through the full pullback, and check the soundtrack at both ends.

The Short states $7 and 8.5 minutes for its example. That is not a guaranteed cost or turnaround. A saved follow-up quoted 195 credits per 720p render and approximately 390 for the initial version plus flamingo revision; it was an estimate, not an inspected billing receipt. Check current prices and count revisions.

## The takeaway

Create the larger scene change first. Revise only the float using the accepted result, then restore the permitted soundtrack and inspect the finished video.

## Workflow resources


## Sources and limitations

Prepared from the saved 12 September 2026 production conversation and original Short. Website setup checked against Higgsfield’s official Genjutsu guide on 2 November 2026: https://higgsfield.ai/creator-hub/help-center/tools/how-do-i-use-genjutsu . Prompts are reusable adaptations; exact motion matching and the headline cost/time are not guaranteed.

[Original video](https://www.youtube.com/watch?v=vbE3LluHQXs)
