Lovely RouterDocs
Get a key
Playground

Images

Generate and edit pictures from the dashboard, combine up to 16 reference images, keep a saved character's face across images, and download or re-edit the results.

Open Images from the sidebar's Playground group. It sends image jobs: each image queues on the server, keeps going if you leave the page, and is billed to your balance like any playground call.

Settings#

The settings sit in a panel on the left. On a phone, the model picker stays at the top and the rest opens with Options.

SettingWhat it does
ModelThe image model. A model on our own GPUs shows an Asleep, Waking up or Warm badge under it.
Channel groupAuto — best available, or a channel group of your choice.
SizeOnly the sizes the model can make, from 1024 × 1024 up to 4K on models that support it, or Let the model choose.
Qualityauto, low, medium or high. GPT Image 2.5 adds xhigh and max. auto leaves it to the model.
CharacterOptional. See Use a character. Only shown for models that take reference images.
Reference imagesOptional. See Edit and combine images.

At the bottom of the panel you see the model's price for your account: a flat price per image, any size or quality, or a token price with a minimum per image. With more than one image per prompt, the total for all of them is shown too. Current prices are on /models and in Images pricing.

Generate#

Describe the image in the prompt box and press Generate, or Ctrl+Enter (Cmd+Enter on a Mac). The prompt takes up to 8,000 characters, and its Write for me / Improve helper can write or polish it for you (a chat call, billed to your balance). With an empty gallery, Try a prompt offers a few examples to start from.

The buttons next to Generate set how many images you get from one prompt: 1, ×2, ×3 or ×4. The button then reads Generate 2, Generate 3 and so on. The playground remembers your choice. From code, n goes up to 10 (Several images).

Images take 40 seconds to 2 minutes each.

Edit and combine images#

Add images under Reference images: click Add images, or drop files on it. You can add up to 16 PNG, JPEG or WebP images, up to 20 MB each. Remove one with the × on its corner.

  • One image: the prompt edits it. Describe the change, for example "make it night with a full moon".
  • Several images: the model combines them. Say in the prompt what to take from each, for example "the jacket from the second image".

With reference images the button reads Edit image, or Make 2 edits and so on with a higher count. The request becomes an edit, as with POST /v1/images/edits.

Use a character#

To keep the same person across images, click Add a character in the panel, or Characters in the header. The Characters dialog lists your saved characters and lets you make a new one, from a photo or a description (see Characters & scenes). Pick one and press Use this character.

The character's sheet then goes along as the first reference image, and the prompt is told to keep their face, hair and build. Change picks another one; the × removes it. If you switch to a model that can't take reference images, the character is removed and a note says why.

Results#

Your most recent images appear below the prompt, newest first, one tile per image. They are kept for 7 days.

  • A queued image shows its place in line and a Cancel button. Once it is generating it can't be cancelled.
  • A finished tile shows the model, the size, the quality, how long it took and what it cost. Click the image to open it full size.
  • Download saves the file. Edit this image puts it in Reference images, as the only one, and clears the prompt so you can describe the next change.
  • Under the prompt, the copy button copies it and the arrow button puts it back in the prompt box. A Character label marks images made with a character.
  • A failed image shows the reason. You are not charged for it.

Up to 20 images per account can be queued or generating at once (How requests are queued).

Get the code#

View code shows the cURL command for the current settings: POST /v1/images/generations, or POST /v1/images/edits when you have reference images. Replace $LOVELY_API_KEY with a key from API keys. For sizes, edits, jobs and limits from code, see Images.