10 Ways to Prompt Photorealistic Cars in Midjourney
Before you jump to conclusions – Midjourney or another AI tool is not a replacement for true photography, and even to a larger extent, CGI. This new wave of AI tools whether it be Midjourney, Dall-E, ChatGPT, Microsoft OpenAI or even Google’s Bard just represents the beginning and can be very useful.
I’m a digital artist at heart and I’ve been doodling since I was old enough to hold a pencil. My mind is constantly filled with ideas that are lost to myself and the world if I don’t do anything with them. Midjourney serves two purposes for me as a creative: a tool to quickly process my ideas into visual representations, and to help build concepts for myself and clients that save a lot of time. I might post these concepts to social media, but I don’t claim them to be photos and I don’t hide behind them being renders from AI, and neither should you.
It's my personal recommendation that Midjourney and any other AI tool is used to help you improve your own workflow, design concepts, or a just to have fun with. Below you will find examples of prompts I’ve used to generate near photorealistic.
Sign Up for Midjourney here: https://www.midjourney.com
Download / Create a Discord login here: https://discord.com
Update - March 21, 2023 - Version 5 Midjourney is available
Midjourney has updated to Version 5, which has major improvements upon the last version. My initial review is much better accuracy, color and resolution. If the system is not rendering in version 5 for you, you can add “—v 5” at the end of you prompt. Example below.
Prompt: a dark purple Nissan Skyline GT - R at sunset with cherry blossom trees, wind, windy, shot on a Sony DSLR, 50mm lens f/ 2. 8, ultra detailed ,--v 5
Example #1. Red BMW E30
Prompt: red e30 BMW M3 at sunset, photography, real, realistic, 8k, bbs, snow, mountains, photorealistic, ultra detailed, --v 4
Tip: If you subscribe to Midjourney, create yourself a private Discord channel and invite the Midjourney Bot to your channel. This will give you a clutter free space to imagine all you want without it getting lost in the mix. Note: This image was generated after using several image prompts from my Cinnabar E30 photo set and rerolled after without the image prompts. Learn more about image prompts here: https://docs.midjourney.com/docs/image-prompts
Example #2. Toyota 4Runner in the Mountains
Prompt: black toyota 4runner in the forest, mount rainier, muddy, photography, real, realistic, 8k, bbs, snow, mountains, photorealistic, ultra detailed, --v 4
Tip: Use Image Prompts to train Midjourney. When you post an image link in the prompt, you’re giving the AI tool more data to work with. You can use multiple images in each prompt for better learning. It’s been my experience that it doesn’t always create photorealistic images this way, but continuing to feed it this information, it improves with each pass.
Example #3 Midnight Purple R34 Skyline
Prompt: dark purple nissan skyline gtr in a neon city at night, shot with a sony mirrorless, 35mm, photography, cinematic, anti-aliasing, CGI, --v 4
Tip: Rerolls! Don’t be shy about rerolling to get more options. Some of the examples here were rerolled several times to iron out setting, style, and features. Select a variation that you like, and you can even reroll those until you land on a final selection.
Example #4 Aston Martin DB9 in London
Prompt: photo of an aston martin db9, in london, at night, shot on a Sony mirrorless camera, DSLR, 50mm lens f/2.8, ultra detailed, 8k, --v 4
Tip: Use real-world details, equipment, and settings to better define that “realistic” look. Use keywords for actual locations like “London” and equipment like cameras or lenses such as “Sony Alpha A7R” and “50mm Lens f/2.8” to provide the AI tool with more information on how to process the render.
Example #5 Porsche 911 in Miami
Prompt: photo of a porsche 911 in miami, sunset, neon, shot on a Sony mirrorless camera, DSLR, 50mm lens f/2.8, ultra detailed, 8k, --v 4
Tip: Weight your prompt keywords by importance. If you’re looking to use an image prompt and want your render to be a close replication of that, put it first, or near the beginning. Otherwise weight the prompt keywords based on what’s most important to you. In the example above, I weight the location and lighting (via sunset) as the priority. The “neon” keyword provides additional coloring and the remaining keywords help apply the photorealistic style of the render.
Example #6 Mazda RX-7 (or something like it?)
Prompt: a photo of a 90s JDM Mazda RX-7 FD3S at night, neon lights, rain, particles, shot on a Sony mirrorless, DSLR, --v 4
Tip: Not every render is going to be an exact representation. When attempting to render a Mazda RX-7 FD3S, you can see that its clearly pulling in elements from a Miata and even other makes and models (maybe even a Dodge Viper?). If you’re a car person, or sharing these with car enthusiasts, they will notice so be careful what you attempt to pass.
Example #7 Honda NSX in Kyoto Cherry Blossoms
Prompt: Black Honda NSX in kyoto at sunset, rain, cherry blossoms, reflection, side profile, sunset, photography, 8k, realistic, shot on sony mirrorless camera, ultra detailed, photorealistic, --v 4
Tip: Don’t forget the mood! Provide context to the tool by giving it a mood to work with. Some great examples are time of day (Sunset, Night), Time of the year (Fall, Winter, etc), Background lighting like (Neon, Mist, Fog) are all good indicators of setting the mood.
Example #8 Ford Raptor in the dunes
Prompt: photo of a Ford F-150 Raptor racing in a sand dune, sunset, motion, shot on a Sony mirrorless camera, DSLR, 85mm lens f/8, ultra detailed, 8k, --v 4
Tip: Avoid text, logos and emblems whenever possible. As powerful as Midjourney and other AI tools are – they lack the ability to do text and/or characters. If you’re using image prompts for sources, expect that big grills like “TOYOTA” or “FORD” will look like characters from a Sci-Fi movie. In the sample above, I used a combination of a render + photoshop to make the grill correct.
Example #9 Audi R8
Prompt: photo of an Audi R8, sunset, shot on Sony mirrorless camera, DSLR, 35mm lens f/2.8, ultra detailed, 8k, --v 4
Tip: Avoid upscaling resolution (for now). The new standard for rending prompts is “- - v 4” which is the latest and greatest rendering quality at the time of this post (typically 1024x1024px). Midjourney offers beta upscaling and others such as “- - testp” to attempt photorealistic images in a higher resolution. Use at your own discretion, but it’s my own experience and opinion to avoid it for now.
Example #10 Ferrari LaFerrari
Prompt: photo of a Ferrari LaFerrari in Rome, sunrise, shot on a Sony Mirrorless camera, DSLR, 85mm lens f/4, ultra detailed, 8k, --v 4
Tip: Prepare yourself to have to edit in Photoshop to finalize. Because of all the reasons above, sometimes the AI tool just won’t give you exactly what you want. You might need to clone out a badly rendered emblem, fix a wheel that a reroll or variation will not.
Go nuts and experiment! That’s the beauty of this tool and you should not be afraid to learn and use it. If anything, use it just to get a better understanding of what it can and can’t do.
To see more recent renderings - follow me on Instagram! @mackeydesigns
