Quick Links
Generating text inside an image isn’t a big deal with version 6 of Midjourney. You will still encounter incorrect spelling, but you may remedy this using variations and choosing the right keywords. Let’s show you how to do it.
Step 1: Select a Midjourney Model That Produces Text
We tested three different Midjourney models to see how well they could produce text inside an image and the results show a clear winner: Midjourney v6.
Version 6 marks the first Midjourney model that has reliable text-generating capabilities, without using a long workaround to achieve it.
You can check what model you are using by typing/settingsin the text box, and then hittingEnteron the keyboard. Click on the drop-down menu and change the model to v6, or a newer model; use whichever is the latest release.
If you’re new to the platform and need more guidance, see our guide onhow to get started with Midjourney.
Step 2: Add Text to Your Midjourney Prompt
Words that you want to appear in your image must be put inside quotation marks. For the prompt to work well, it should also be somewhat logical, explaining where the words should appear and how they are written.
In our example of a retro billboard painted with the words “1970s World Fair”, we’ve described where the words should appear (the billboard) and how they are written (painted).
Step 3: Use the Right Keywords to Generate Images With Text
There are plenty of ways to describe writing. Painted, printed, embossed, stamped, graffitied, scratched, inscribed, you name it. Use these adjectives and others like them to describe exactly what you want.
Take the examples below:
We used two almost identical prompts, except one uses the adjective “written” and the other doesn’t. This small difference changed the entire meaning of the image produced.
If you want to generate text on its own, which is not necessarily a part of a scene, the keyword you need isTypography Design. It’s a great trick to create logos or design material that you may use for other projects.
Step 4: How to Get Good Text Results in Midjourney
Don’t get hung up on bad results because you can fix them by spinning out a few variations. In the example below, we used the variation button two more times to eventually arrive at an image that was just about the same as the original, but with the correct spelling.
You can do the same by hitting one of theVbuttons labeled 1 to 4, corresponding to the four images in the set.
Using the Remix Feature
You can take this method one step further using the remix function. Turning this setting on will let you edit the prompt when you hit the variation button, giving you more control over the image and making it easier for you to change the text in the image.
Type the command/prefer remixin the text box, then hit enter twice. You will receive a message saying “Remix mode turned on”, confirming your settings.
The next time you use theVbuttons to create variations, a dialog box will appear with the original prompt where you are free to make changes. Swap out the text if you like, and if you need to refer back to what keywords work best, you can review step 3.
As a general rule, one to two words will work best. Common phrases you are likely to see in the context of the image will also work, as opposed to something nonsensical. For example, a sign reading “Honeymoon Motel” had more success than “Crater Comforts”.
Save these techniques for later and try them out on other projects likecreating an AI-self portrait.
While it’s disappointing to see spelling mistakes and doubled-up words, you can clean up the results relatively fast by creating variations, using the remix function, and getting the keywords right. When youcompare the best AI generators, it’s these features that make Midjourney stand out.
So, without too much work, and with far fewer workarounds than in previous models, you can create some pretty good text in Midjourney.
