Introduction

A practical guide to using the GPT Image 1.5 model in our Media Generator: options, costs, prompt tips, and best practices.

GPT Image 1.5 is OpenAI's latest image generation model, tuned for better efficiency and quality. In our platform, it powers:

• Photo generation — Images from text prompts
• Image projects — Batch jobs with multiple prompts
• Chat agents — In-chat image generation (when configured)
• Reference-based generation — Using product or style images as input

Use it for general-purpose image creation when you want strong quality and consistency with our simplest workflow.

Where to Use GPT Image 1.5

| Location | How | |----------|-----| | Media Generator | Tools → Media Generator → Create Job → choose "GPT Image 1.5" in the model dropdown | | Image projects | Create a project, add prompts, and select GPT Image 1.5 as the model | | Chat | Use an agent that includes image generation and has GPT Image 1.5 selected as the image model | | LLM preferences | General Settings → LLM Preference → set GPT Image 1.5 as the default image model |

Platform Options

Aspect ratio

| Option | Label | Output size | Best for | |--------|-------|-------------|----------| | 1:1 | Square | 1024×1024 | Icons, thumbnails, social grid | | 16:9 | Landscape | 1536×1024 | Banners, headers, wide layouts | | 9:16 | Portrait | 1024×1536 | Stories, mobile-first content | | 4:3 | Standard | 1024×1024 | Classic photo crop | | 3:4 | Photo | 1024×1536 | Portrait photos, cards |

Default:

• 1:1 (square)

Output format

• PNG — Lossless, good for graphics and transparency needs

Batch size (images per request)

• 1–12 images per generation
• Generating multiple images at once is more efficient than separate requests

Quality

• Auto — Platform default, balances quality and speed
• Quality can be adjusted via API parameters when supported

Reference images

• Optional — No reference required for pure text-to-image
• Up to 4 reference images when editing or style-transferring
• Accepted formats: PNG, JPG, WEBP
• Max file size: 10 MB per file
• Max files: 10 uploads per request

Reference images are useful for:

• Product photos with consistent styling
• Keeping a character or object consistent
• Style transfer from example images

Costs (credits)

GPT Image 1.5 uses flat per-image pricing in platform credits:

• Cost is per generated image
• Exact credit amount is shown in the model selector in the Media Generator
• Batch jobs: cost = number of images × per-image cost
• Costs may be adjusted; always refer to the UI for current rates

Prompt guidelines

Do

• Be specific: subject, style, composition, lighting, mood
• Use clear, natural language; the model understands longer prompts
• Mention composition: “centered,” “wide angle,” “close-up”
• Specify style: “photorealistic,” “watercolor,” “minimalist”
• Include lighting: “soft daylight,” “studio lighting,” “golden hour”
• Add context: setting, background, atmosphere

Example

> A professional product photo of a ceramic coffee mug on a light wooden table, soft window light from the left, white background, minimalist style, commercial quality

Avoid

• Overly vague descriptions (“nice image,” “pretty picture”)
• Conflicting instructions in the same prompt
• Extremely long prompts; clarity usually beats length
• Prompts that violate content or safety rules

Good practices

1. Start simple, then refine

Begin with a short prompt. If the result is off, add detail for:

• Subject
• Style
• Lighting
• Composition
• Mood

2. Use reference images for consistency

• Upload 1–4 reference images when you need:

- Consistent product look - Specific style - Same character or object across images

• Keep references clear and well-lit

3. Choose the right aspect ratio

• 16:9 — Banners, headers, desktop layouts
• 9:16 — Stories, mobile, vertical feeds
• 1:1 — Thumbnails, grids, square social posts

4. Batch when possible

• Use batch projects or multiple images per request
• More efficient than many single-image runs

5. Prompt enhancement (projects)

• If enabled, the platform can enhance prompts before generation
• Useful when you have rough ideas; review and tweak before running

Limits and constraints

| Constraint | Value | |------------|-------| | Max prompt length | 32,000 characters | | Reference images | Up to 4 per generation | | Rate limit | 10 requests per minute | | Max images per request | 4 (batch) |

Troubleshooting

| Issue | What to try | |-------|-------------| | Wrong style or composition | Add explicit style and composition terms | | Inconsistent results | Use reference images or more specific prompts | | Rate limit errors | Space out requests; avoid burst usage | | Credits exhausted | Check balance; add or upgrade plan as needed | | Model not available | Confirm OpenAI API key is configured and GPT Image 1.5 is enabled |