GPT88 Agent Image Studio Guide

Entry points and positioning

This is not a chat-first product. It is a task-oriented image workstation for uploading reference assets, selecting generation modes, shaping prompts, and exporting finished visuals.

Recommended entry points
  Image studio: https://agent.gpt88.cc/
  Console / API keys: https://gpt88.cc

Common related areas
  Image Studio: /image-studio
  Prompt case library: /xsct-cases
  Usage / balance: console dashboard
  API keys: console keys

Best-fit scenarios

• Ecommerce hero images and marketplace cover images.
• Scene images for product lifestyle presentation.
• Model images for fashion, beauty, accessories, and product trials.
• Detail-page assets, material closeups, and sales-point graphics.
• Background replacement, inpainting, outpainting, and upscaling.

Preparation

• Sign in and confirm balance.
• Create an API key if you want to reuse results in automation later.
• Prepare product photos, reference images, brand color direction, and fixed product constraints.
• Confirm image-model access such as gpt-image-2 or Gemini image models.

Recommended workflow

The most stable path is draft first, scale later. Do not start with expensive large-size batch jobs.

1. Sign in and confirm usable balance
2. Open https://agent.gpt88.cc/
3. Choose the task mode: main image, scene image, model image, detail asset, background removal, inpaint, extend, or upscale
4. Upload the product image or reference image
5. Choose model, ratio, resolution, format, and quality
6. Generate one small draft first
7. Lock the direction, then move to 2K / 4K or batch generation
8. Download results and verify usage in the console

Example prompt skeleton for an ecommerce hero image:

Subject: one luxury skincare serum bottle, glass body, silver pump
Scene: minimal white bathroom counter, soft natural light, clean premium look
Composition: centered product, keep top headline area free, suitable for ecommerce hero image
Style: commercial photography, realistic materials, premium detail
Limits: no text, no watermark, no logo, do not change bottle structure

Mode overview

• main: ecommerce hero and key product image
• scene: lifestyle or environment scene image
• model: model-worn or person-centered presentation
• detail: detail-page material or selling-point asset
• remove-background: clean cutout or simplified background
• background: replace or standardize the background system
• inpaint: local repair or partial image replacement
• extend: outpainting and ratio adaptation
• upscale: final resolution enhancement

Models and specs

The workstation exposes image models according to account permission. In practice, usegpt-image-2 when you want a stable API-oriented flow, and use the visual workstation when you want manual drafting and fast directional validation.

• Common resolutions: 1K, 2K, 4K.
• Common ratios: 1:1, 3:4, 4:3, 2:3, 3:2, 9:16, 16:9.
• Common formats: png, webp, jpeg.

Batch generation advice

Before batch generation:

1. One sample image has already confirmed style, composition, and product consistency
2. The prompt explicitly locks non-changeable elements such as bottle shape, color, and package structure
3. Target ratios are listed clearly: 1:1, 3:4, 4:3, 16:9, 9:16
4. Validate in 1K first, then move to 2K / 4K
5. Batch task names are clear enough for download, filtering, and reuse

Relation to API and Codex skill flows

• Open Agent Image Studio for manual drafting and visual review.
• Image Generation API for direct programmatic use.
• Codex gpt-image-2 Skill for file-producing Codex automation.
• Models for model IDs and protocol references.

Cost and quality advice

• Use low resolution and single-image drafts to confirm direction before moving to 2K / 4K.
• Do not mix text layout and base image generation into one request when quality matters.
• For ecommerce, product consistency matters more than style experimentation.
• Review prompt lists before batch runs so you do not multiply the wrong direction.