Build a Reel from three separately generated shots: encounter, contact, and memory. Use the shot-specific English prompts and editing plan below, then match the contact poses on your timeline. This manual workflow differs from generating one complete video on this site.
This existing example illustrates the transition. The timing below is an editing plan, not a measurement of this clip or a guarantee from a video model.
Reference clip · 15.0s · 480×854
Input photos and model are unverified. This reference does not promise the current generator’s result.
In CapCut: import the three clips, arrange them in this order, set a 9:16 canvas, and split at contact. Match screen sides and arm positions, then align music to the actual memory cut. Replay the exported file to check faces, framing, and sound.
Timeline target
Shot and edit
English prompt for an external video tool
0–4 s
Establish the encounter. Keep the same two references and leave room for both faces in a vertical frame.
Use Photo 1 as the adult survivor and Photo 2 as the beloved person. Establish a quiet encounter on a ruined street at dusk. Keep both recognizable faces visible, Photo 1 on screen left and Photo 2 on screen right. No dialogue, text, gore, or extra characters.
4–8 s
Recognition and contact. Choose a usable contact frame before generating the next shot.
Continue with the same references, clothes, screen sides and camera distance. The survivor recognizes the beloved, lowers their defenses and approaches slowly. End at the first gentle embrace, with both shoulders and hands clearly visible. No face swap or new characters.
8–15 s
Continue the pose in a healthy memory. Trim to matching body positions rather than forcing an exact timestamp.
Continue the final contact pose from the previous shot in a warm seaside memory. Use the same two identities, now both healthy. Keep camera distance, screen sides and arm positions consistent at the opening. Hold the embrace, then show a quiet shared smile. No text or dialogue.
In each shot prompt, Photo 1 is the survivor and Photo 2 is the beloved person. Keep the same references and screen sides across all three clips. Save a usable contact frame from the second shot and use it as a continuity reference for the memory if your external tool supports that input.
02
Budget the beats before generating
For a 15-second plan, reserve time for the encounter, recognition, approach, contact, and memory. Avoid asking a short clip to contain a long chase, several locations, and many flashbacks. If your external tool supports a different duration, adjust every time window so the story finishes within that duration.
03
Specify continuity at the handoff
Describe where each character stands, the direction of movement, the camera angle, and the arm position at contact. Ask the warm shot to continue that pose. These instructions help define your intention; they cannot guarantee a frame-perfect transition or exact likeness.
04
Review generated material as separate decisions
First check identity and role order, then natural anatomy, then whether the contact can connect to the memory. Changing music will not fix a swapped face. Keep the usable take, note the specific failure, and revise the relevant instruction instead of adding more unrelated details.
05
Move from generation to the editing pass
Download the source material and keep an untouched copy. Assemble the sequence, trim the handoff, and add the final audio and caption only after the visual story works. The manual Reels editing guide focuses on those timeline decisions; this page focuses on preparing usable AI shots.
Common questions
Does the prompt guarantee the same faces?
No. Reference quality and model behavior still affect identity. Review every generated scene and replace shots that drift too far.
Can I paste this into any AI app?
Use a video tool that supports the required reference inputs and duration. Adapt unsupported timing or split the storyboard into shorter shots.
Does your generator need manual prompting?
No. It supplies a 15-second storyboard from your photos and settings. The copyable prompt is for a manual workflow.
Should I upscale before checking the story?
Check the original identities and movement first. More pixels do not repair incorrect roles or a disconnected embrace.