AI Prompt Guide
Q: How do I make realistic photographs? How do I get a clear photographic style image? How do I make my image look like a photo? The quickest route to photographic quality is to cite real sources of photography, such as magazines (e.g.
by Vogue), famous photographers (e.g.
by Robert Frank), and photography types such as
cinematic still shot or
portrait photography or
stock photography. The best algorithm for photography is
--testp which you can set as a parameter at the end of the prompt, or select in
/settings. Avoid style phrases like
hyperrealistic, hyperrealism, realism, photorealistic, photorealism, which are used by artists to describe painterly styles of art that mimic photos. Avoid style phrases like
rendered, Octane Render, Unreal Engine, hd, 4k, 6k, 8k and other terms from CGI unless you want your photo to have a waxy look with depth-of-field and/or tilt-shift effects.
Phrases to add to your prompt that MIGHT help change the camera zoom and angle:
- Wide-Angle Shot – Pull the camera back from subject
- Ultra-Wide Angle – Pull the camera very far back from subject
- Satellite View – Pull the camera into the stratosphere
- Eye-Level Shot – Camera even with head of figure
- Far-Shot Angle – Figure looks very small/distant
- Medium-Shot Angle – Looking up at figure from knee level (child gazes up at parent)
- Ground-Shot Angle – Looking down at figure from knee level (parent gazes down at child)
- Low-Angle Shot – On ground, looking up at figure (puppy gazes up at human)
- Full-Shot Angle – Your target in full view
- Full-Body Shot – Head-to-toe view, like full-shot angle
- Glamour Shot – Different zooms, but always flattering to figure
- Cinematic Still Shot – Emphasis on context/background
These phrases are more influential the closer to the beginning of the prompt they appear.
How do I get a full body picture? How do I get the whole person in the image? Crop, cropping?
How can I turn the head of my model, make them look down, or look up, or look to the side?
Give the model an emotion, attitude, or even an activity that is correlated to the position you want to see. Example words that change the position of the shape or body include:
You may also try “–no facing camera” as a negating prompt
1100+ artists trained by MJ (v3 engine) along with multiple example generations for each. Add more via the WIP tab:
Midjourney Artist Influence Reference Sheets
Several hundred artists separated by specialty, with links to their work:
CLIP Ranked Artists
Hundreds of artists, ranked by how well CLIP-based models match the artist’s style, and therefore relevant for test/testp use. Most helpful: examples of artwork can be seen next to generated images with direct links to more examples:
Please extend a Big Thank You to Clarinet on Midjourney server whatnostop#6700
Here are some ti[ps on referencing artists in prompts
Mix and match unlike artists: Not happy with your prompt? It might be the artist (or artists) you chose. Craig Mullins might be perfect for you 99/100 times, but maybe he doesn’t play nicely with Gerald Brom or Frank Frazetta—experiment a LOT, happy accidents are real! 3 artists are almost always better than 1: Every artist specializes, some are better at faces, some are better at environments, etc.; hedge your bets by including 3 artists whose strengths/weaknesses compliment one another or cover a wider range (ex: Dean Cornwell, John Howe, Frank Lloyd Wright) MJ loves some artists more than others: Norman Rockwell, Greg Rutkowski, Studio Ghibli seem to produce consistently solid results Safe or go-to artists: If you’re not sure where to start or who to choose, here are 3 artists that always work well in MJ: Artgerm, ArtStation, DeviantArt (the latter two are artist collectives representing a wide range of skills)