Tags: video-generation identity-consistency smartphone-selfie Video Template: Text Parameterized Input: Twoidentity_reference_images (image) Input: city (text) Output: Paired_city_selfie_videos (video) bytedance-seedance-2-5 bytedance-seedance-2

Paired Nighttime City Selfie Videos with Identity Consistency

Create two 30-second handheld smartphone selfie videos in the same city at night: one featuring the woman in her reference image and one featuring the man in his. Preserve each person’s identity while deliberately retaining imperfect phone-camera artifacts.

By Threads — Published on

Example results

Download video result 1
Download video result 2

Prompt template

Create two contrasting 30-second smartphone selfie videos set in the same city at night: one featuring the woman in the first reference image on a nighttime city ride, and one featuring the man in the second reference image on a nighttime city ride.

At the beginning of each prompt, include this identity instruction: “Preserve [ref_image]’s face, hairstyle, skin tone, and body proportions, but ignore the reference image’s clothing, pose, and background. Lock identity, not the frame, so the scene can be rewritten.”

Use the same city, ${city}, for both videos. Make the footage feel like rough, authentic handheld phone selfies rather than polished cinematic video. Intentionally include occasional missed autofocus, hand shake, overexposure, and compression noise. Keep these imperfections natural and believable.

Parameters

${city} city Required

Comments — 0

Related prompts