gpt-image-2 · 3:4 · 2K · medium quality.
Generation model
select
default:"gpt-image-2"
The AI model used to generate all five pose images.
- GPT Image 2 (
gpt-image-2) — default model, optimized for photorealistic fashion imagery following OpenAI’s virtual clothing try-on guidance - Reve 2.1 Image (
reve-2.1-image) — alternative model for experimentation
Aspect ratio
select
default:"3:4"
Controls the output image dimensions. Choose based on your primary publishing destination.
Resolution
select
default:"2K"
Output resolution.
- 2K — high-resolution output suitable for marketplace listings and print
- 1K — lower resolution for quick drafts and internal review
Generation quality
select
default:"medium"
Controls the level of detail, generation time, and cost per image.
- Low — fastest draft quality, lowest cost; use for rapid iteration
- Medium — balanced default; suitable for most catalog and marketplace requirements
- High — maximum detail and accuracy; highest cost; use for hero images or premium campaigns
Model identity
select
Describes the model used for every pose in the session. All five frames are generated to match this identity.
To lock a specific face rather than an ethnicity description, upload a Model face lock image in the Product references section.
Changing model identity marks the current analysis stale and triggers an automatic re-analysis before generation can start, because the model description is part of how Gemini plans the shoot.
Background style
select
Determines the backdrop, environment, and lighting direction used across all five poses.
You can also type a custom scene description directly in the Scene direction field in the Analysis & Pose Plan panel after analysis runs.
Changing background style marks the current analysis stale — Gemini re-analyzes with the new scene direction before generation can start.
Pose QA
boolean
default:"false"
When enabled, Gemini runs a consistency validation check on every generated pose image before it is marked complete. QA checks the generated image against the Product Identity Profile and the approved consistency anchor.
- Enabled — each pose goes through an extra Gemini QA pass; poses that fail are automatically retried with a minimum 30-second backoff
- Disabled — poses are delivered without automatic validation; QA status is shown as “Unverified” in History
Styling plan editor
After analysis completes and the result is current, the Footwear, jewellery & styling panel appears below the Output settings. Gemini proposes a specific footwear, jewellery, and accessory plan based on your product photos and style references. The styling plan includes:- Footwear style and color
- Necklace, earrings, bangles, and ring descriptions
- Handbag or clutch
- Additional accessories (dupatta, belt, hair accessories)
- Nail color