Skip to main content
After you upload the required front and back product references, Gemini automatically starts a vision analysis to build two structured profiles that drive the entire photoshoot: a Product Identity Profile and a Creative Direction Profile. Together, these profiles lock every visual decision — garment fidelity, model styling, lighting, and scene — before a single image is generated.

​
Automatic analysis trigger

Gemini analysis starts automatically a moment after you finish uploading your references or making any meaningful input change. You do not need to click anything — analysis runs in the background as soon as the required references are present. The Scene & styling panel shows the current state:
  • Analyzing… — Gemini is actively processing your references
  • Locked — analysis is current and the shoot is ready to generate
  • Stale — rebuilding automatically — an input changed; generation is locked while the automatic rebuild runs
You can also click Analyze now (or Analyze again if a previous result exists) to trigger a manual analysis, or Improve Pose Plan to force a fresh Gemini call that regenerates the entire pose plan even when the current inputs haven’t changed.

​
Product Identity Profile

The Product Identity Profile captures everything about the physical garment that must be preserved exactly in every generated image.

​
Creative Direction Profile

The Creative Direction Profile drives scene, lighting, and editorial feel. All five poses share the same scene so the final shoot looks like it was captured in a single session.

​
Editing Gemini’s output

You can override Gemini’s default scene and garment read in two editable fields: Scene direction — pre-filled with a condensed summary of background style, studio environment, lighting, and mood. Edit this field to change the backdrop, lighting, or mood. For example, type ‘Warm golden-hour terrace, diffused natural light, soft editorial mood’ to replace the AI’s studio default. Garment summary — pre-filled with category, main color, fabric, and key invariant details. Edit this field to correct any misread detail, such as a fabric type Gemini couldn’t identify from the image. Both edits are sent as director’s notes on the next analysis. Gemini re-derives the Product Identity Profile, Creative Direction Profile, and five-pose plan around your correction. Your edits are never silently overwritten by a fresh analysis — they persist until you clear them.
Editing either field marks the analysis stale immediately and starts an automatic rebuild. Because your edits are included in the analysis cache key, repeating the same edit you’ve already sent — or reverting to the previous value — hits the existing 30-day cache rather than triggering a new paid Gemini call.

​
The five-pose plan

Gemini creates a pose plan tailored to your specific garment. The five standard poses cover every angle needed for marketplace listing and social media. Each pose card shows its camera angle, framing, primary reference, body position, hand placement, expression, and highlighted details — all derived by Gemini from your specific garment.

​
Editing individual poses

Click any pose card in the Poses section to expand its full detail view. The Generation direction textarea at the bottom of the expanded card contains the prompt that will be sent to GPT Image 2 for that frame. Edit it directly to add corrections, emphasize a specific detail, or restrict an element. For example, for Pose 3 you might add: ‘Back side should not have hanging or latkan elements. Preserve the plain uploaded back construction exactly.‘

​
Improving the pose plan

Click Improve Pose Plan in the Scene & styling panel to regenerate the entire five-pose plan with a fresh Gemini call. This bypasses the cache and produces a new pose plan around the same references and settings. Use this when the AI’s initial plan doesn’t suit the garment’s style, the editorial direction, or your target marketplace.
All five poses must be enabled and have non-empty generation direction prompts before the Generate button becomes active. If any pose has an empty prompt, the generation is blocked.

​
Saree-specific analysis

For sarees, Gemini builds an extended Saree Truth Profile and a Saree Drape Plan in addition to the standard Product Identity and Creative Direction profiles:
  • Saree Truth Profile covers body fabric, border details, pallu artwork, pleat zone behavior, blouse construction, and fabric physics separately.
  • Saree Drape Plan specifies drape family, shoulder side, waist tuck, pallu spread, fall direction, border visibility, and blouse visibility for each pose.
If Gemini detects that the saree analysis is incomplete — for example, the pallu evidence is missing or ambiguous — the Scene & styling panel shows an alert: ‘Stored saree analysis is incomplete or outdated. Reanalyse the product references before generation.’ Generation remains locked until you upload the missing references and rerun analysis.

​
Stale analysis

Any time you change a reference image, category, model identity, background style, or any other input, the analysis is marked stale because it no longer matches the current set of inputs. A stale analysis means:
  • The pose plan shown may not match your current inputs
  • The Generate button is disabled
  • Re-analysis starts automatically a moment after the change
You can also click Analyze again to rebuild immediately without waiting for the automatic trigger.