Pindeck can generate new images derived from any existing image in your library. Generated images are linked to their source via a parentImageId, forming a parent-child lineage you can navigate and filter.
The generation model is fal.ai Nano Banana Pro (fal-ai/nano-banana-pro/edit), an image-editing model that takes a source image and a short prompt and returns one or more new images.
How to trigger variation generation
- Hover over any image card in the gallery.
- Click the magic wand (Generate) icon button.
- Select Variations from the dropdown menu.
- Configure the options in the Generate Variations modal and click Generate.
- Click any image to open its detail modal.
- Click the Generate button in the action row.
- Select Variations.
- Configure and submit.
After uploading images, a Generate Variations dialog appears automatically. Configure the mode, aspect ratio, and count, then click Generate to queue variations for all uploaded images at once.
Modification modes
Choose a mode to control the type of variation Pindeck generates. Each mode maps to a different prompt sent to the model.
| Mode | What it generates |
|---|
| Shot variation | A different camera angle or framing of the same subject. Automatically cycles through 16 shot types when no custom direction is given. |
| B-roll | The same location or environment with no people visible — establishing shots, interiors, objects. |
| Action shot | A dramatic or kinetic narrative moment featuring the same person. |
| Style variation | The same person in a different scene, outfit, or location. Same face and likeness, new context. |
| Subtle variation | A later moment in the same scene — same person and setting, slightly different action or position. |
| Coverage | Detail or object shots within the same environment. No people. Useful for cutaway inserts. |
The 16 shot types (used by Shot variation)
When you run Shot variation without a custom direction, Pindeck cycles through these shot types in order, assigning one per generated image:
close-up · extreme close-up · close-up profile · medium close-up · macro of the hands · profile · from behind · over-the-shoulder · medium shot · wide shot · extreme wide shot · low angle shot · high angle shot · bird’s eye view · dutch angle · point of view shot
Group-aware prompting
The mode prompts adapt their language based on the Group assigned to the original image:
- Music Video — prompts are prefixed with “Later in the music video.”
- Commercial — prompts are prefixed with “Later in the commercial.”
- Film / TV Series / Web Series — prompts use scene language (“Later in the scene.”) for action shots; shot-variation prompts are unchanged.
This means the same “Shot variation” mode produces subtly different prompts depending on the production context of the source image.
Generation settings
| Setting | Options | Default |
|---|
| Variation count | 1–12 | 1 |
| Aspect ratio | 16:9, 9:16, 1:1, 4:3, 3:4 | 16:9 |
| Custom direction | Free-text prompt addition | (blank — automatic) |
Parent-child lineage
Every generated image stores the parentImageId of the image it was derived from. This creates a navigable chain:
- The Table view shows a Parent column. Click the parent title to open the parent image’s detail modal.
- The Originals only toggle in the gallery and table view hides all images that have a
parentImageId, showing only direct uploads.
For images that originated from Discord, the lineage depth is capped at 12 levels. Variations of Discord-origin images beyond depth 12 will not be auto-queued.
Reviewing generated images
Generated variations land in a pending status. They appear in the AI Suggestions section of the Upload tab while they wait for review. From there:
- Keep — approves the variation; it moves to the Review & Finalize queue and eventually to the gallery.
- Discard — permanently deletes the generated image.
Use the Originals only toggle in the gallery to hide all AI-generated children and focus on your uploaded source images.