GPT Image 1.5 - Image Generation & Editing

Generate new images or edit existing ones using OpenAI's GPT Image 1.5 model via the Images API.

Usage

Run the script using absolute path (do NOT cd to skill directory first):

Generate new image:

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "your image description" --filename "output.png" [--quality low|medium|high] [--size 1024x1024|1024x1536|1536x1024|auto] [--background transparent|opaque|auto] [--output-format png|webp|jpeg] [--output-compression 0-100] [--n 1-4] [--api-key KEY]

Generate multiple variations:

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "your image description" --filename "output.png" --n 4 --quality high

Edit existing image (without mask - full image edit):

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "editing instructions" --filename "output.png" --input-image "path/to/input.png" [--quality low|medium|high] [--input-fidelity high|low] [--output-format png|webp|jpeg] [--api-key KEY]

Edit with multiple reference images (compositing/style transfer):

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "combine the subject from first image with the style of second image" --filename "output.png" --input-image "subject.png" --input-image "style-ref.png" [--input-fidelity high] [--api-key KEY]

Edit existing image (with mask - precise inpainting):

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "what to put in masked area" --filename "output.png" --input-image "path/to/input.png" --mask "path/to/mask.png" [--api-key KEY]

Important: Always run from the user's current working directory so images are saved where the user is working, not in the skill directory.

Parameters

Quality Options

low - Fastest generation, lower quality
medium (default) - Balanced quality and speed
high - Best quality, slower generation. Use for dense text, detailed layouts, and high-fidelity output

Map user requests:

No mention of quality -> medium
"quick", "fast", "draft" -> low
"high quality", "best", "detailed", "high-res" -> high

Size Options

auto (default) - Model decides best dimensions based on prompt
1024x1024 - Square format
1024x1536 - Portrait format
1536x1024 - Landscape format

Map user requests:

No mention of size -> auto
"square" -> 1024x1024
"portrait", "vertical", "tall" -> 1024x1536
"landscape", "horizontal", "wide" -> 1536x1024

Background Options (generation only)

auto (default) - Model decides
transparent - Transparent background (requires png or webp output format)
opaque - Solid background

Output Format Options

png (default) - Lossless, supports transparency
webp - Smaller file size, supports transparency, good for web
jpeg - Smallest file size, no transparency

Map user requests:

No mention of format -> png
"small file", "compressed", "web" -> webp
"jpg", "jpeg", "photo" -> jpeg
"transparent" -> png (or webp)

Output Compression (webp/jpeg only)

Range: 0-100 (default: 100 = no compression)
Lower values = smaller file size, lower quality
Only applies to webp and jpeg output formats
Ignored for png

Number of Images (n)

Range: 1-4 (default: 1)
Generates multiple variations in a single API call
When n > 1, files are numbered: output-1.png, output-2.png, etc.

Map user requests:

"variations", "options", "alternatives" -> --n 4
"a couple" -> --n 2
No mention -> 1

Input Fidelity (editing only)

high - Preserves fine details, facial identity, textures from input image(s)
low - More creative freedom, less strict preservation

Map user requests:

Face/portrait editing, identity preservation -> --input-fidelity high
"keep the face", "preserve likeness", "maintain details" -> --input-fidelity high
Style transfer with loose reference -> omit (let model decide)

API Key

The script checks for API key in this order:

--api-key argument (use if user provided key in chat)
OPENAI_API_KEY environment variable

If neither is available, the script exits with an error message.

Filename Generation

Generate filenames with the pattern: yyyy-mm-dd-hh-mm-ss-name.{ext}

Format: {timestamp}-{descriptive-name}.{ext}

Timestamp: Current date/time in format yyyy-mm-dd-hh-mm-ss (24-hour format)
Name: Descriptive lowercase text with hyphens
Extension: Match the --output-format (.png, .webp, .jpg)
Keep the descriptive part concise (1-5 words typically)
Use context from user's prompt or conversation
If unclear, use random identifier (e.g., x9k2, a7b3)

Examples:

Prompt "A serene Japanese garden" -> 2025-12-17-14-23-05-japanese-garden.png
Prompt "sunset over mountains" (webp) -> 2025-12-17-15-30-12-sunset-mountains.webp
Prompt "create a photo of a robot" (jpeg) -> 2025-12-17-16-45-33-robot.jpg

Image Editing

All editing uses the Images API (images.edit endpoint) with gpt-image-1.5.

Without Mask (Full Image Edit)

When the user wants to modify an existing image without specifying exact regions:

Use --input-image parameter with the path to the image
The prompt should contain editing instructions (e.g., "make the sky more dramatic", "change to cartoon style")
A fully transparent mask is auto-generated, allowing the model to edit the entire image

With Mask (Precise Inpainting)

When the user wants to edit specific regions:

Use --input-image parameter with the path to the image
Use --mask parameter with a PNG mask file
The mask should have transparent areas (alpha=0) where edits should occur
The prompt describes what should appear in the masked region

Multi-Image Input (Compositing & Style Transfer)

When the user wants to combine elements from multiple images or apply a style from one image to another:

Use --input-image multiple times (e.g., --input-image "subject.png" --input-image "style.png")
Reference images by order in the prompt (e.g., "apply the style of the second image to the subject in the first image")
Use --input-fidelity high when identity/detail preservation matters

Common editing tasks: add/remove elements, change style, adjust colors, replace backgrounds, composite images, style transfer, virtual try-on, text translation in images.

Prompt Handling

For generation: Pass user's image description as-is to --prompt. Only rework if clearly insufficient.

For editing: Pass editing instructions in --prompt (e.g., "add a rainbow in the sky", "make it look like a watercolor painting")

For multi-image: Reference images by position (e.g., "combine the person from the first image with the background of the second image")

Preserve user's creative intent in all cases.

Output

Saves image to current directory (or specified path if filename includes directory)
File extension is automatically matched to --output-format
For n > 1, files are numbered: name-1.ext, name-2.ext, etc.
Script outputs the full path(s) to the generated image(s)
Do not read the image back - just inform the user of the saved path(s)

Examples

Generate new image:

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "A serene Japanese garden with cherry blossoms" --filename "2025-12-17-14-23-05-japanese-garden.png" --quality high --size 1536x1024

Generate with transparent background:

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "A cute cartoon cat mascot" --filename "2025-12-17-14-25-30-cat-mascot.png" --background transparent --quality high

Generate multiple variations as compressed webp:

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "Minimalist logo for a coffee shop" --filename "2025-12-17-14-26-00-coffee-logo.webp" --n 4 --output-format webp --output-compression 80 --quality high

Edit existing image (full image):

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "make the sky more dramatic with storm clouds" --filename "2025-12-17-14-27-00-dramatic-sky.png" --input-image "original-photo.jpg" --quality high

Edit with identity preservation:

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "change the outfit to a red dress" --filename "2025-12-17-14-28-00-red-dress.png" --input-image "portrait.png" --input-fidelity high --quality high

Composite from multiple images:

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "place the person from the first image into the beach scene from the second image" --filename "2025-12-17-14-29-00-beach-composite.png" --input-image "person.png" --input-image "beach.png" --input-fidelity high

Edit with mask (inpainting):

uv run ~/.claude/skills/gpt-image-1-5/scripts/generate_image.py --prompt "a flamingo swimming" --filename "2025-12-17-14-30-00-lounge-flamingo.png" --input-image "lounge.png" --mask "mask.png"