The current state of AI tools for Designers in 2023
There’s a lot of hype around all sorts of AI tools lately. Most public sources chase sensationalism to get views, but the reality and usability of these tools are often quite different. This article is therefore intended to help you gain an understanding of the really useful tools out there that are worth exploring and to understand the limitations of the other ones that, in their current state, you’ll just be wasting your time with.
There is no doubt that the advent of artificial intelligence will fundamentally change our approach to work in the coming years. It’s not just the design community that is now going through the initial excitement of what AI tools can do, but few people yet have successfully incorporated these tools into their daily workflows. Yet even in the current state of these tools, which are only just revealing the true possibilities of the very near future, we can benefit from them and fundamentally improve efficiency.
1. Image generation
Tools such as DALL-E 2, Stable Difusion, or Midjourney can now generate everything from photorealistic images to specifically stylized artwork. The more specific your instructions, the better results you get. Ironically, the only limit is your imagination. Therefore, it is no problem to generate images for social networks, blogs, or, for example, illustrations for presentations.
The problem arises when you want to work with the images further. Most tools will generate 4 variations of a low-resolution image and you can then upscale your favorite. That is, generate it at a higher resolution. However, even this is not enough in some cases, and the AI often runs out of creativity when it has to figure out the details when zooming in. Then you’re forced to run the image through other AI tools specialized just for enlarging images and filling in missing details.
Further complications can arise when you want to generate multiple images with the same styling. For example, for visually unified illustrations for branding or digital product.
You can only achieve this with a very long and specific text prompt. And even then, the desired result is not guaranteed and you end up preferring to reach for verified, hand-made photo banks and illustration collections.
Midjourney
Midjourney was most useful to us so far. It handles all kinds of images and art styles perfectly. Moreover, it is not a problem to request, for example, a specific camera type, lens, and lighting, a desired aspect ratio, or a specific art style. The downside is that it is only accessible via Discord and thus not a very user-friendly experience, with the need to remember at least basic text commands.
+ Creativity and a high level of quality
+ Detailed specification of the prompt
+ You can try the first few dozen images for free. After that 8$ per month
- Need to use Discord
- User unfriendliness
DALL-E 2
DALL-E 2 produces the most consistent results while generating rather simple objects against a plain background, or conversely photorealistic close-up shots of people. Its greatest feature is a generation of missing parts of the image, so that you can for example take a close-up of the character and conjure up the entire room a character is standing in.
+ Most consistent results
+ Can fill in missing parts of existing images
- Must specify exact requirements, otherwise little creativity
- Only 20 free images to try out. After that 0.02$ per image
Stable Diffusion
Stable Diffusion approaches Midjourney in terms of quality of output and user (un)friendliness, although the results are not at the same level. The advantage is the open-source nature of the project, thanks to which there is a large fan base around Stable Diffusion, plus the tool is free to use. But we prefer to stick with the proven Midjourney.
+ Great creativity
+ It’s free
+ Large community (open-source)
- User unfriendliness
2. Generating (not only) text content
If you work in any creative industry, you can’t avoid working with text. Whether you’re coming up with texts for websites and apps, creating marketing materials, pitch decks, presentations, or product descriptions, you can always use a helper to spark your creativity, help you improve the content you’ve already written, or write some parts for you. Here I dare to say that Chat GPT-3 is the only tool you will need.
There are plenty of apps to help you with specific types of content (emails, newsletters, blog articles, or video/podcast scripts), but Chat GPT-3 will help you with anything without having to pay for individual tools that focus on one thing. Additionally, Chat-GPT-3 is currently free with the prospect of a paid plan coming soon, but trust me, it still shouldn’t be missing from your arsenal. We use it daily not only for proofreading texts, filling in blind spots in articles, and getting comprehensive information on complex topics but also for generating pieces of code or translations.
Chat GPT-3
+ Generate grammatically correct text in many languages
+ Perfect understanding of the context
+ Broad knowledge of all disciplines including very peripheral topics
+ Possibility of adding very useful extensions and additions
+ So far it is completely free
- May produce incorrect or inappropriate answers
- Often unavailable due to traffic overload
- Knowledge of information only until 2021
Another AI discipline related to the text is language translation. Here, we can’t fault DeepL, which is free, can translate entire documents, and achieves better results compared to Google Translate. Thanks to AI-powered translations, it understands the context well and the translations are thus mostly usable without any proofreading.
3. Editing photos and graphics
As far as photo editing is concerned, we can mention remove.bg, which removes the background of almost any object with one click and handles hair and other complicated shapes well. It’s enough for a quick check of an idea, but in the end, you often still need to use Photoshop to clean up the details.
And while we’re on the subject of Photoshop, did you know that Adobe has been adding highly advanced AI-powered features to almost all of its Creative Cloud tools for years? Relatively recently, for example, Photoshop added Neural Filters to help photos enhance their details, remove various blemishes, transfer style from one photo to another, or better remove backgrounds from objects. In After Effects, you can remove unwanted objects from a video, or generate missing parts of an image, and Lightroom will go through and tag all your photos for you so you can quickly find them without the need for manual tagging. For Creative Cloud users, I recommend reviewing the descriptions of each feature on Adobe Sensei.
If you “just” need to enlarge a poor-quality photo and fill in missing details, then reach for one of the many web-based tools that do this very well. We most often use Let’s Enhance, which offers free editing for the first 10 images and then costs $ 9 a month.
4. Working with audio
There are many text-to-speech tools, but most virtual speakers still have a very robotic voice. However, we did manage to find one of the few exceptions! ElevenLabs generates the most realistic voice I’ve heard so far, and I wouldn’t be afraid to use it for commercial marketing materials and voiceovers. You can even clone an existing person’s voice by uploading few minutes of his/her voice recording and just like that you can generate realistic audio from a text prompt. Another tool that can also do this quite well is Descript, a powerful tool for creating not only podcasts, combining several useful functions for working with text and audio. Be sure to also check out Adobe Podcasts, which includes similar features but takes audio work to a higher level.
ElevenLabs
+ Totally realistic voice
+ Possibility to clone real voice
+ Free up to 10,000 characters per month, $5 per month thereafter
+ Web app, no need to install anything
- If it starts to become widely used, everyone will recognize it
Another tool that we’ve incredibly appreciated is Enhance Speech from already mentioned Adobe Podcasts, which can take a low-quality audio recording of the spoken word and enhance it to near-studio quality. But it can be used in other ways, too. We often record audio of conferences and the spoken word recorded on a phone from the back row is usually barely understandable. With one click, you get a near-perfect recording cleaned of all noise that you won’t be embarrassed to share with your colleagues.
Enhance Speech
+ Perfect studio-quality audio even from a low-quality microphone
+ Very easy to use
+ It’s free
+ Web app, no need to install anything
- Voice often changes slightly, as if another person is speaking
What tools were not included but are worth mentioning?
🧐 VisualEyes
VisualEyes simulates eye-tracking studies and preference tests with a 93% accurate predictive technology. This tool could save days of user testing and give graphic designers instant feedback. Unfortunately, we haven’t had the space to test this tool yet.
🤩 Runway
Runway is the most advanced AI video editing tool. It can remove backgrounds or unwanted objects from video, set object motion tracking with one click, automatically edit a movie to the beat of the music, add subtitles, or remove noise from audio. We haven’t had a chance to try the tool ourselves that much, but we know from other people that it works really well and we’re looking forward to editing some videos when we have a chance.
🤔 Synthesia
Synthesia — probably the most famous platform where you can choose from a range of photorealistic virtual persons who will read the required texts for you. The result can be a frighteningly realistic product presentation narrated by a non-existent person. We haven’t had a use for this tool yet because we want our content to be authentic and based on real people on the team. But if you need a professional presentation and don’t have a camera or a suitable person on hand, why not…
👎 Uizard
There has been a lot of hype around Uizard for quite some time. It has several AI features, probably the most interesting of which is converting hand-drawn wireframes into functional designs. On paper, a great idea that has been toyed with for years by Airbnb e.g. In reality, however, the tool has never generated anything usable for us and has always been a complete waste of time. And so were Uizard’s other AI features. Thumbs down for us.
Conclusion
There is no AI tool that can flawlessly solve complex problems for us. Rather, only certain parts of the design process. Think of AI as just another tool to help us get to the result we need, without having to waste time on tedious, repetitive processes. Alternatively, it can help us break a creative block when we lack inspiration.
The downside of the current era of early adopters of AI is that many companies are developing their own AI tool, which is often just a sort of technical proof of concept, which can only handle small tasks and often does not work reliably. If we then want to incorporate AI for our work on a larger scale, we have to pay large subscriptions for a lot of tools that we don’t even need that often.
Therefore, the future might be dominated by companies that can combine multiple tools into one and offer them to experts in various creative disciplines at a reasonable price, or gradually upgrade features of already existing applications to a new AI-powered version. Just as Adobe is currently doing with its Sensei. In any case, we look forward to such a future and are ready for it. And you? 🙂