Living portrait: a subtle breath of motion
Creates a subtle portrait loop with a blink and small natural movements from your photo. Upload your starting image, paste this prompt into Gemini or Grok, change the parts in brackets to suit you, and generate.
- Written for
- Gemini, Grok, Other video models
- You'll need
- A starting photo
- Status
- Tested 9 Oct 2026 in Grok Imagine video (9 Oct 2026; model id not exposed)
- Test settings
- Exact default prompt text, with the bracket set to its published default, was submitted on 9 Oct 2026 in Grok Imagine video. Synthetic AI-generated adult portrait used as the starting frame (/media/living-portrait-reference-tested-20261009.webp). The prompt asks for a 5-second loop; this generator's shortest clip is 6 seconds. First result kept, 720p. The reference has no leaves, so background motion was not requested by the scene. Re-encoded to H.264 with no creative edits. A still frame from the clip is the example image. ChatGPT and Gemini were not used.
- Variability
- Blink timing, hair movement and how closely the last frame matches the first can vary. This is not a perfect loop and not a guarantee for another face.
Prompt
Animate the uploaded portrait into a subtle 5-second loop. Use the photo as the first-frame reference. Keep the head position and framing nearly fixed. Add one gentle blink near the middle and slight breathing; move loose hair or clothing only if visible in the photo. Use leaves swaying slightly out of focus only when that element exists in the photo; otherwise keep the background still. Lock the camera and exposure with no push-in or cuts. Bring the last frame as close as possible to the first. Preserve recognisable facial features, natural skin tone, hair and clothing as closely as the model permits. Avoid speech, lip movement, face morphing, extra blinking, warped hands, flicker and sudden motion.
Customise
For example: leaves swaying slightly out of focus.
Fields start with illustrative values; replace them with your own facts. The prompt updates as you type. Copy prompt copies your version.
How to use it
- Choose a video-generation tool that accepts the required input. Availability depends on the tool and your account.
- Upload your starting photo.
- Paste the prompt and generate. Make two or three takes and keep the best.
How to use this prompt
- Choose your starting photo. Get permission from people who can be identified, and use an app that accepts an image-to-video input.
- Upload the photo in the external app, copy the prompt with your details, paste it and generate the video.
- Play the whole video and check faces, objects, motion and the change from the first frame. The still shown here is not a real video test.
Good to know
- Face details and the loop can still vary. Use a consenting subject and inspect the full clip before sharing. The motion plate is illustrative.
- Use reference photos you have the right to use; get consent from identifiable people.
