Write a text to image prompt
Lead with medium and light, then the subject. Leave out what you do not care about.
Midjourney AI Image Generator
AI Image
Midjourney is an AI image generator for text to image. After a frame comes back you can upscale it or inpaint a described region instead of generating the whole picture again. It reacts strongly to medium, lighting and style words.
How to generate an image with Midjourney
Generate, then upscale or inpaint
When a frame is back, upscale raises resolution on that frame. Inpaint replaces a described region. Those are actions on the result, not separate model names.
Set the ratio, then generate
Choose aspect ratio and how many images to return before you run.
What Midjourney is for
Text to image, then upscale and inpaint
Describe the picture, pick a frame, then upscale to raise resolution on that exact frame, or inpaint to replace a described region while leaving the rest alone. Zoom and pan are follow-up actions on a finished frame.
How to phrase a Midjourney prompt
Name the medium and the light before describing the subject in detail. Long lists of unrelated adjectives tend to average each other out.
Several options per run
Ask for more than one image per run and only upscale the frame that is closest. That is cheaper than enlarging everything and choosing afterwards.
Questions people ask about Midjourney
What is Midjourney?
Midjourney is an AI image model for text to image. After a picture is back you can upscale it, inpaint a region, zoom or pan without generating a new frame from scratch.
What is Midjourney upscale and inpaint?
Upscale raises resolution on the frame you already chose. Inpaint replaces a described region and leaves the rest of the picture alone.
How should I write a Midjourney prompt?
Name the medium and the lighting first, then the subject. Short, specific style words work better than a long list of unrelated adjectives.
Does Midjourney do image to image?
Follow-up actions run on a finished Midjourney frame: upscale, inpaint, zoom and pan. Start with a text to image prompt, then edit that result in place.