How to take full control over your AI generations

A step-by-step guide to manage every detail of your generated photo and get AI-generated content that precisely fits your design.

The internet is loaded with image generators for any taste. You can type the prompt and get what the AI is generating for you in minutes based on the query. When you generate images just for fun and memes, you don’t look for much fidelity in details. So you reach the result in just a few tries. However, now AI is coming to the professional content field more and more, and sometimes, the picture you need is very particular. How to get the exact result?

Well, you can master the art of prompting, learning how to get the best shot with the prompt. On our blog, we have covered some basic tips to prompt more efficiently. On the other hand, you can save time and use a generator that has your back already with preset settings.

Check out the video guide all learn from the step-by-step tutorial in the article.

The purpose of the generated images

Let’s set the design task where we need to use AI-generated content as part of the composition.

Here, our task is to design the UI for a mobile app. It is an online course platform. We defined the frames, colors, and layout for two options of the course tab.

Now, to present the result, we need to show an example of the content. We know that the app will be targeting the Indonesian market. Let’s ask ChatGPT to generate the UX copy:

This way the mockup looks already more related to the case. But something is still missing: we need photos of our course expert. The UI requires three different photos of the same person, and this person should look reliable for the app’s future audience. Quite a task! Let’s try to generate the photos.

Setting the person’s identity

We open Human Generator. It gives us a random image to start with, but we reset all the settings to start from scratch.

In Human Generator, the prompt is just a tool to pinpoint the details. For the main characteristics of our future humans, we use dropdowns and sliders in the settings panel. We chose an Indonesian middle-aged woman and set the main color of her clothes as brown. For the start, we set the standing pose.

Looks good! The next step is to determine the outfit. Since our woman is an expert, let’s dress her more to the formal side. We go with wide-leg pants and oxfords.

Pro tip: if you don’t like the very first result given, try to click “Generate variation” button. However, if you want to keep almost everything the same, don’t generate another variation and just focus on playing with the setting. This is how your results will stay close to the same human type.

The result after setting the exact clothes already looks closer to what we need. Interesting detail: after dressing our model more formally, the generator automatically picked a kind of oriental-styled office background, even though we haven’t set it up yet.

The only detail that seems a bit off is this fashion choice with the basque around the woman’s waist. Let’s fix that by determining the shirt. We also manage the background: since our sample course is about the coffeeshop business, let’s set the background as a vintage room and add “in coffeeshop” to the prompt section.

Here we go! Last but not least, let’s make our lady look more friendly. For this, we put “closed smile” in the prompt and “showing teeth” in the negative prompt section. We do that to avoid a smile becoming too wide, which often can lead to tiny facial twitches.

Voila! As you can see, the result looks almost the same. The only difference is just the smile. It looks like we got our personality to use in the mockup. It took us around just a few minutes and 4 generations.

Same person, different poses

Now we need to generate images for other areas of the mockup. By the legend, it should be the same person, but as if those photos were taken on other days and with a different context. Let’s start with the image in the sitting position. We changed the background set to the studio backdrop and removed the coffeeshop from the prompt. The first result that we got had big earrings, and we put earrings to the negative prompt to exclude them from the result. The best image we got with this pose choice:

The woman looks good, but the thing is that it is a different woman! And our approach is to keep the identity the same. To fix this, we upload our generated human from the previous step to the generator at the “upload face” tab to use the same face traits and update the result. Now we’re good.

The final step is the headshot for the profile pic section. We set the pose as the portrait and played with different neutral backgrounds. The best result so far we got with the white wall background.

Final result

This is it! With roughly 20 minutes, we got three realistic AI-generated photos of the same-looking woman in different poses for our design. Now, the mockup looks the most relevant to its final purpose, and we are ready for the presentation.

By using the Human Generator settings accurately and adding what we needed step-by-step, we were able to control each detail of the result. Just like we have this lady as the royalty-free model in real life in the photoshoot.

Pro tip: if you need even better quality of the face for your design, you can enhance the generated image using the Smart Upscaler. This will give you more contrast to the face traits.

Try yourself

Here is the link to the image generation that we used in this use case. You can click manage it the way you need. You can also find more inspiration in the “Human Templates” section in the top right corner of the generator interface.