Image-to-image editing

If you've been trying to edit an existing image with gpt-image-2 and nothing changes — your reference image is ignored, /v1/images/edits returns 404 — that isn't a bug, it's the model's capability boundary. This page explains why, and gives you two working routes.

Why gpt-image-2 can't edit images#

On this platform gpt-image-2 is a pure text-to-image model: it accepts a prompt and does not consume reference images.

About /v1/images/edits: that was OpenAI's early (DALL·E-2) multipart-only editing endpoint. Here, image-to-image is folded into /v1/images/generations — pass an image array and it's image-to-image, omit it and it's text-to-image. So there is no separate /edits route; models that can edit just use /generations.

Route 1: use a model that supports image-to-image (recommended)#

The Seedream and Nano-Banana families support image-to-image natively on the same /v1/images/generations endpoint — put your reference into the image array and it's done in one call.

Models that can edit#

Reference-image input formats differ slightly per model, so a publicly reachable https URL is the safest bet (every model accepts one). data-URIs are only accepted by the whole Seedream line and nano-banana-2.

The request#

{
  "model": "doubao-seedream-4-5-251128",
  "prompt": "Make the sofa blue, leave everything else unchanged",
  "image": ["https://your-host.com/original.png"],
  "size": "2048x2048"
}

image can also be written as image_urls (equivalent). With several references, the upstream treats the first as the primary edit target and the rest as style or element references.

curl#

curl https://jiuye.zsopc.com/v1/images/generations \
  -H "Authorization: Bearer $PLATFORM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seedream-4-5-251128",
    "prompt": "Make the sofa blue, leave everything else unchanged",
    "image": ["https://your-host.com/original.png"],
    "size": "2048x2048"
  }'

The response#

The same OpenAI Images shape as text-to-image, with data[].url holding the edited image:

{
  "model": "doubao-seedream-4-5-251128",
  "created": 1765432100,
  "data": [{ "url": "https://...", "size": "2048x2048" }],
  "usage": { "generated_images": 1 }
}

Route 2: GPT-5.x chat with the image_generation tool#

If you'd rather stay in the GPT pipeline, use the built-in image_generation tool on a gpt-5.x chat model: pass the reference as a multimodal message, and the model looks at the image and redraws it.

The request#

curl https://jiuye.zsopc.com/v1/chat/completions \
  -H "Authorization: Bearer $PLATFORM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "stream": true,
    "tools": [{ "type": "image_generation" }],
    "messages": [{
      "role": "user",
      "content": [
        { "type": "text", "text": "Make the sofa in this image blue, leave the rest alone" },
        { "type": "image_url", "image_url": { "url": "https://your-host.com/original.png" } }
      ]
    }]
  }'

The response#

The edited image arrives inline in the streamed body, as markdown image syntax:

data: {"choices":[{"delta":{"content":"![](https://...generated...)"}}]}

Regex ![](url) out of the accumulated delta.content to get the link.

This route bills by token, not per image: an image_generation turn carries a fixed overhead of roughly 2,300 input tokens, plus the vision tokens for any reference image (roughly doubling it), with very few output tokens. It suits complex edits that need the model to understand the image before changing it.

Which route to pick#

For ordinary edits — recolouring, swapping elements, style transfer — prefer route 1: faster, cheaper and predictable.

Limitations#

• Don't use gpt-image-2 for editing — it's text-to-image only and the image parameter does nothing.
• Don't call /v1/images/edits — the endpoint doesn't exist here (editing goes through /generations with an image array).
• Image URLs returned by route 1 are upstream pre-signed links that expire in about 24 hours — download and store them promptly.
• If your key has an allowed_models allowlist, add the editing model to it first or you'll get a 403.
• Full image / video / music endpoint documentation: Image / video / music APIs.