Generate Live Photo
AI-generatedSummary
Generate an iOS-style live photo video from a starting image, optionally interpolating toward an end image and applying a motion prompt. Poll Gemini until generation finishes, process the video with ffmpeg into a short motion-and-freeze clip, and upload it to S3.
Inputs
- Video Model — Choose Veo 3.1 Fast (default) for faster generation or Veo 3.1 for higher quality.
- Image URL (required) — URL of the image used as the starting frame; the node fetches it before sending it to Gemini.
- Motion Description (Optional) — Describe the desired motion. If left blank, the node uses a default prompt for focus, lighting, and ambient shadow changes.
- End Frame URL (Optional) — Optional second image used as the final frame. When provided, Gemini interpolates between the frames and generates an 8-second source video; otherwise it generates a 4-second source video.
- Aspect Ratio — Select landscape (16:9, default) or portrait (9:16). The processed video is scaled and padded to 1280×720 or 720×1280, respectively.
- S3 Bucket Name (required) — S3 bucket to receive the processed MP4. S3 credentials are also required.
- S3 Public Domain — Optional domain used to construct the returned video URL instead of the S3 endpoint URL.
- Additional Options — Set the status polling interval (default 10 seconds; UI range 5–60) and maximum wait time (default 30 minutes; UI range 5–60).
Output shape
A single n8n item per input item, with JSON metadata and no video binary data.
JSON includes videoUrl, fileName, model, type, resolution, aspectRatio, duration, usedFramesToVideo, imageUrl, endFrameUrl, prompt, and usedDefaultPrompt. The uploaded processed MP4 is 3 seconds long at 24 fps, although the returned duration field is the fixed string "4s". Processing requires ffmpeg on the n8n host; generation or upload errors fail the item, or return an {"error": "..."} JSON object when Continue On Fail is enabled.