Skip to main content
Sora Image (sora_image) and gpt-4o-image are not in LaoZhang API’s current public model catalog. This page is kept for reference by existing integrations. To edit images, use GPT Image 2.5, Nano Banana 2, or FLUX image editing; other models are compared in Choose an image generation API.

Model Overview

The legacy Sora Image editing route submitted image URLs and edit instructions through the chat completions interface (/v1/chat/completions), then parsed the edited image URL from the reply content.

Interface Capabilities

  • Image inputs: Supports one or more image URLs as input
  • Text instructions: Describe edits in natural language, such as style transfer, element changes, or multi-image fusion
  • Chat completions interface: Submit image and text content through /v1/chat/completions
  • Result parsing: Extract the edited image URL from response text

Feature Comparison

Request Examples

Basic Example - Single Image Edit

Advanced Example - Multi-Image Fusion

Editing Scenarios

1. Style Transformation

2. Smart Background Replacement

3. Object Editing

4. Color and Lighting Adjustment

Best Practices

1. Precise Editing Instructions

2. Multi-Image Fusion Tips

3. Batch Processing

Important Notes

  1. Model Selection:
    • gpt-4o-image: Legacy image-generation model ID used by existing integrations
    • sora_image: Legacy Sora Image route
    • Neither ID is in the current public model catalog
  2. Image Input:
    • Supports online image URLs
    • Ensure images are publicly accessible
    • Recommended max size: 20MB
  3. Multi-Image Processing:
    • Can include multiple images in one request
    • Add all images to content array
    • Describe fusion/editing intent clearly
  4. Result Extraction:
    • Results returned in Markdown format
    • Extract URLs using regex
    • Download edited images promptly

FAQ

Q: Can I edit local images?

A: Upload local images to a public hosting service first, then use the URL. Or use base64 encoding (may increase costs).

Q: How many images can I process at once?

A: No strict limit, but 2-5 images per request is recommended for best results and performance.

Q: Can I use both Sora and GPT-4o Image?

A: Both were legacy model IDs on the chat completions route, and neither is in the current public model catalog. For new editing work, use GPT Image 2.5.

Q: Are the edits reversible?

A: No, each edit generates a new image. Save originals if you need to revert changes.

Q: What’s the quality of edited images?

A: Results depend on prompt clarity and the quality of the original image.