Skip to main content

Step 3: Generation Options

This step controls what the final video looks and sounds like. The sidebar calls it Options.

Step 3: Generation OptionsStep 3: Generation Options

Output settings

Two selects are always visible:

  • Resolution: Auto (match clips) (default), 4K, 1080p, 720p. Auto picks the tier of your source clips; 720p renders fastest, 4K needs the source material and the patience.
  • Output Format: MP4 (H.264), MP4 (H.265), MOV (H.264), MOV (H.265), MOV (ProRes). The default follows output.codec and output.format in your config — MP4 (H.264) out of the box.

Advanced options

A collapsed expansion under the two selects:

  • Orientation: Auto (detect from clips) (default), Landscape (16:9), Portrait (9:16), Square (1:1). When the source and target aspect ratios differ, the whole frame is kept and the leftover space is filled — clips are never cropped, see face-aware framing.
  • Scaling Mode: Blur background (default), Letterbox (black bars). Starts on whatever defaults.scale_mode says, so the wizard and --scale-mode produce the same video.
  • Transition Style: Smart (mix of fades & cuts) (default), Crossfade, Cut, None. Smart picks per cut point: crossfade for slow scenes, hard cut for fast action.
  • Add date overlay, Caption clips with their place, and Keep intermediate files checkboxes. The place caption is the wizard's equivalent of the CLI's --add-place, and needs GPS on the clips to show anything.
  • When photos are enabled: "Photos enabled (N found)" and a Photo duration (seconds) field (1–10).
Advanced generation optionsAdvanced generation options

Title

Title and Subtitle fields pre-filled with the pipeline's suggestion (LLM-generated when content analysis is on, template otherwise), a Language select (auto, en, fr, de, es, it, nl, pt, ja, ko) and a Regenerate button. Trip memories also show chips with the detected trip and map mode. See Title Screens & Maps.

Music

The Background music select has up to four entries:

  • None: original clip audio only.
  • Upload file: a Select music file picker (.mp3, .m4a, .wav) and a Music volume slider.
  • Bundled: a royalty-free track from the installed music package, chosen to match the memory's mood and the photo cut cadence. Only listed when that package is present — offering it otherwise would resolve to silence, which is what None is for. Picking it skips AI generation even when a generator is configured.
  • AI Generated: only listed when ace_step.enabled or musicgen.enabled is set in your config, and then it is the default. Shows the volume slider and a Generate Music button so you can hear the track before rendering; Regenerate Music tries again. See Audio & Music.

Summary

Stat cards for Clips, Duration, Resolution and Music, then Back to Clip Review / Next: Preview & Export.

Hardware acceleration is not shown here; it is auto-detected at encode time. Run immich-memories hardware or immich-memories preflight to see what your machine offers.