Image Generation

Image generation always uses the same models (currently: gemini-3.1-flash-image-preview). Choosing a model in this case refers to the LLM that parses the text and constructs a prompt passed onwards to the image generation model.

Speed and credit usage includes the full run with the results provided.

Summary

3.1 Pro

3 Flash

3.1 Flash Lite

2.5 Flash

Speed

34.4s

40.4s

29.1s

28.7

Credits

11

8

7

8

Based on the results we think 3.1 Flash Lite or 2.5 Flash are the best first try models. They seem to be able to create prompts feed well into the image generation model. They take shorter time and are cost effective.

Gemini 3.1 Pro

Results

  • Speed: 34.4s

  • Credits: 11

Image

Gemini 3 Flash

Results

  • Speed: 40.4s

  • Credits: 8

Image

Gemini 3.1 Flash Lite

Results

  • Speed: 29.1s

  • Credits: 7

Image

Gemini 2.5 Flash

Results

  • Speed: 28.7s

  • Credits: 8

Image

More fun at the table, less fuss around it

The tavern falls quiet. You spent all week on this villain: the scarred voice, the secret letter, the slow reveal in the final act. Your players have other plans. They have just set fire to the only bridge out of town, the paladin is deep in an argument with a goat, and the rogue has quietly pocketed the letter that was supposed to unravel everything. Somewhere across the river, your villain stands alone in the rain, waiting for an entrance that will never come. The bard leans back, grinning. Four faces turn to you.

What do
you do?

Start Free

30 days free · no card needed