"AI hug video" is one of the most searched AI video ideas, usually for the same reason: two people who are not in the same place any more, and one photograph where they are.
Here is how those clips are actually made, without pretending the app has a feature it doesn't.
What Rigel gives you
Rigel generates video from a still image plus a scene description. There is no hug preset, no template gallery entry called "hug." What you have is a prompt field, and the model's ability to generate plausible human motion from the photo you supply. That is enough — but only if the source image and the prompt do their jobs.
Start with the right photo
This is where most attempts succeed or fail.
- Use one photo containing both people. A picture where they already stand or sit together is far more reliable than trying to conjure a second person.
- Leave space around them. Tight head-and-shoulders crops leave nowhere for arms to travel. Half-body or wider works much better.
- Prefer a clear, even light. Mixed light sources make the contact point between two bodies fall apart.
- Keep hands visible and simple. Hands are still the weakest area in this generation of models.
Prompt recipes that work
Describe the movement, the camera and the light — in that order. Keep the action small and physical.
- "They turn toward each other and embrace gently, arms around shoulders, slow natural movement, camera holds still, soft warm indoor light."
- "He steps in and hugs her, she closes her eyes and smiles, slow motion, slight handheld sway, late afternoon backlight."
- "Slow push-in as the two of them lean together into a hug, subtle head movement, shallow depth of field, evening window light."
What to avoid: "emotional reunion," "heartfelt moment," and other mood words with no physical content. The model cannot render a feeling; it can render an arm moving believably.
When the photo is old
Most of these clips start from an old print. The preparation matters: flat, even capture, cropped to the people, best available copy. The full checklist is on the animate old photos page, and the underlying generation workflow is described on the photo to video page.
If you want the subject to speak as well as move, talking photo covers lip-sync from a script.
Set your expectations honestly
Two-person contact is one of the hardest things to generate. Expect to try more than once, expect wider framings to work better than close ones, and expect that some photos simply will not produce a convincing embrace. When it works, it works because the source image already contained most of the answer.
And the rule that matters more than any prompt: only use photos of people who have agreed, or of your own family. Generating intimate footage of someone who never consented is not a technical question.
Try it
Rigel is free to download on iOS 16.6+ and Android and includes starter credits, so the first attempt costs nothing. Credit packs and a subscription are available in-app, and each generation shows its credit cost before you confirm it.