Command for generating images

[simonw]

Given GPT-4o native images are now available in an API, I think it's time to have a llm openai image-generate command or similar.

https://openai.com/index/image-generation-api/

[simonw]

It can use the saved API key. Should it log the results to a database? Including the generated images?

I think so. I think it should duplicate and obey the --no-log option. I'm thinking the images go in a folder in user_dir() and their filenames and prompts are stored in a database table that's created in logs.db - or the database specified by a --database option.

[simonw]

• https://platform.openai.com/docs/guides/image-generation
• https://platform.openai.com/docs/api-reference/images

import base64
from openai import OpenAI
client = OpenAI()

img = client.images.generate(
    model="gpt-image-1",
    prompt="A cute baby sea otter",
    n=1,
    size="1024x1024"
)

image_bytes = base64.b64decode(img.data[0].b64_json)
with open("output.png", "wb") as f:
    f.write(image_bytes)

Even more interesting is the edit endpoint:

result = client.images.edit(
    model="gpt-image-1",
    image=[
      open("body-lotion.png", "rb"),
      open("bath-bomb.png", "rb"),
      open("incense-kit.png", "rb"),
      open("soap.png", "rb"),
    ],
    prompt="Generate a photorealistic image of a gift basket on a white background labeled 'Relax & Unwind' with a ribbon and handwriting-like font, containing all the items in the reference pictures"
)

For image docs say:

The image(s) to edit. Must be a supported image file or an array of images. For gpt-image-1, each image should be a png, webp, or jpg file less than 25MB. For dall-e-2, you can only provide one image, and it should be a square png file less than 4MB.

[simonw]

More from the docs:

n: integer or null. Optional, Defaults to 1.

The number of images to generate. Must be between 1 and 10.

quality: string or null. Optional, Defaults to auto.
The quality of the image that will be generated. high, medium and low are only supported for gpt-image-1. dall-e-2 only supports standard quality. Defaults to auto.

response_format: string or null. Optional, Defaults to url.

The format in which the generated images are returned. Must be one of url or b64_json. URLs are only valid for 60 minutes after the image has been generated. This parameter is only supported for dall-e-2, as gpt-image-1 will always return base64-encoded images.

size: string or null. Optional, Defaults to 1024x1024.

The size of the generated images. Must be one of 1024x1024, 1536x1024 (landscape), 1024x1536 (portrait), or auto (default value) for gpt-image-1, and one of 256x256, 512x512, or 1024x1024 for dall-e-2.

[simonw]

Saving some time:

llm \
  -f github:simonw/llm-openai-plugin \
  -f issue:https://github.com/simonw/llm-openai-plugin/issues/18 \
  -m gemini-2.5-flash-preview-04-17 \
  -s 'Implement a Click command for this based on the arguments in the issue thread'

https://gist.github.com/simonw/6ce46a8eba71adfae59282e031f9e5a3#response

[simonw]

I want a -c option for replying to the first with updates, which actually triggers an edit round with the first generated image as the input.

[talwrii]

Hey. I was just trying to do this now. But the feature does not exist yet! so I guess I'll have to log into the website for now.

Some quick thoughts on use cases. Sometimes I'm probably going to want to put an image directly onto the image for pasting. For this I've used command | xclip -i -selection CLIPBOARD -t image/png before though one could also do this with -t text/uri-list list as well with less risk of breaking terminals by mistake.

[gdbing]

I threw together llm-openai-image plugin yesterday because I hadn't seen this issue and you're usually so prompt about adding support for new models that I assumed you weren't going to do it 😅

I thought about adding some kind of --output file.png arg but instead just directly stream the base 64 image data out of llm and decode and pipe it into a file with |base64 --decode > file.png

llm -m openai/gpt-image-1 \
  -a Maison-Martin-Margiela-HM\ \(4\).jpg \
  -a HM_Margiela_W_16.jpg \
  "A photo of me dressed in these pants and top, in a barren frozen antarctic landscape" \
  -a IMG_2135.jpeg \
  -o size portrait \
  |base64 --decode > margiela-snow-myself.png