Passionate Kiss · skill Open Skill công khai
Passionate Kiss
Passionate Kiss turns your photos into a five-second embrace you can almost feel. Upload one picture of two people, or two portraits, and they step toward each other, wrap their arms around each other, and fall into a continuous, close kiss. The first two seconds are the pull; the rest is the heat, held until the last frame.
ShortDrama
10 2 4
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/3949831/2026-10-13/3ysyx7hj.gif
# Passionate Kiss
Turn the user's photos of people into one short video of them hugging and sharing a passionate kiss. There is no project state. The default length is 5 seconds: the first 2 seconds are only the approach, and from 2 seconds to the end the kiss continues.
The video model receives exactly two inputs: the long English prompt written from `references/kiss_prompt.md`, and the photos the user uploaded this turn. Do not put other reference images, preset assets, or a previous video into `input_urls`.
## When to use
- One photo: the two people in that photo kiss each other.
- Two photos: the person in photo 1 and the person in photo 2 kiss each other, inside the background of one of those two photos.
- The user wrote nothing, the photo contains captions, or the text asks for dancing, walking, dialogue, a wardrobe change, or a new location: still make only this kiss video. If the text states a duration or an orientation, adopt only those two settings.
## What the user sees
Do the work, then say only what the user needs. Do not call `write_todos`. Do not show a task list.
Never mention internal steps or names: reading the skill, `read_skill_file`, `recognize_image`, `generate_video`, model channels, parameters, asset ids, prompt writing, or the long prompt itself. Do not paste URLs. Do not narrate progress ("checking the photos", "writing the prompt", "generating now").
Speak only in these cases, in one or two short sentences:
- No usable photos, or only one person in a single photo: ask for one photo of two people, or two portraits.
- More than two photos: say that only the first two were used.
- A person looks under 21, or no real person is visible: say this video is only for adults, or that a person is needed, and stop.
- The user asked for a clip under 4 seconds: say the approach is the opening portion and the kiss is the rest.
- Success: duration, vertical 9:16 unless they asked for another ratio, and — with two photos — which photo's place is the background.
- Failure: the problem in plain words. Do not claim the video is ready, and do not invent a link.
## Steps
1. Read `references/kiss_prompt.md` with `read_skill_file`. Do not skip this. The action timeline in that file is the source of truth. Do not mention the read.
2. Count the photos uploaded this turn and label them photo 1 and photo 2 in upload order.
- Zero photos: do not call a generation tool. Ask the user for one photo of two people together, or two separate portraits.
- One photo: both people should be in it. If only one person is visible, stop and say a second portrait is needed. Do not invent a partner.
- Two photos: pair the person in photo 1 with the person in photo 2. If a photo shows several people, use the most foreground, most fully visible person. The video background is one photo's location, not a new place. Default to photo 1. If the user names the other photo, use that one. If the default photo has no usable place (solid color, empty studio sweep, or a crop with no setting), use the other photo.
- More than two photos: use only the first two, and say so in the reply.
3. Call `recognize_image` once per photo that will be used. The `prompt` must ask only for what is visible: number of people, whether they look like adults, hair, facial features, skin tone, build, clothing, shoes, accessories, pose, and the place behind them (location, lighting, weather, and whether a real setting is visible). Mark anything unclear as not visible. Do not guess.
4. Safety check, before any `generate_video`:
- If any person looks under 21, is a minor, or age is unclear and the face looks young: stop. Say that this video is only for adults. Do not call a generation tool. Do not mention the skill.
- If no recognizable person is in the photos: stop, explain, and do not generate.
5. Using `references/kiss_prompt.md`, turn the recognition results into one long English prompt. Describe appearance in detail. Use the kiss timeline from that