AI tools have been trending in the last couple of months on social media and in all technology and media news. Although tools based on AI technology are still in early development, people all over the world are already exploring options in ways that AI tools can be used as assistants in marketing, design, sales, product development, and many other fields.
Although AI tools can be used for positive things, speed up processes, and help us out in the workplace, there are also concerns about the ways that AI tools are used and there has been a lot of talk about regulating the usage of AI tools in the future. If you’re interested in that topic, read more about it here.
Some of the best AI tools currently include an AI chatbot, ChatGPT, AI art generators like Midjourney, Dall-e or Stable Diffusion, AI voice generators like Murf and WellSaid Labs, and AI content assistants like Jasper or Copy. With the evolution of these tools, especially OpenAI’s ChatGPT, we should see an explosion of AI tools on the market in the near future with applications in almost all industries (looking forward to Notion AI a lot).
I’ve been using a couple of AI tools in my work for the last 2-3 years but just recently I’ve decided to discover more about all possibilities of art generators and ChatGPT, as the next big movers on the market.
During my research, I decided to create an AI character as a test and as it turned out great, I’ve decided to share the process of creating the character with you.
In the process of creating an AI character, I’ve used multiple tools: Midjourney for art generation, ChatGPT for content inspiration, Play.ht for voiceover generation, Veed for subtitle creation, D-ID.com for character animation and CapCut for final video editing.
Follow the steps mentioned below to create your own AI powered character in 10 minutes (when you get the hang of it).
The first step in creating an AI character is creating a picture of the character in one of the AI tools available such as Midjourney, Dall-e, Stable Diffusion, etc.
My pick for this task was Midjourney as I think that this tool creates more realistic pictures compared to Dall-e which tends to be a bit artsier. Feel free to test out both or even try any other tool and iterate based on the result.
Out of all the tools I’ve used to create the AI generated character, setting up Midjourney turned out to be a bit complicated - you need a Discord account, and you might get some errors when setting everything up. To help you out, here is a short tutorial on how you can create your own Discord account and join the Midjourney server.
If you are not sure what you would like to create or which prompt might be helpful for your AI character, check out PromptHero, a service that searches millions of art images by AI models like Midjourney, DALL-E, Stable Diffusion, etc.
With every picture, you can also see the prompt inserted into the AI that generated that picture, which is quite convenient as you can use that information to create an image in a similar style.
After you have concluded what you would like to create and in which style, head over to one of the “Newcomer rooms” on Midjourney, and in the message bar type your prompt. All prompts should start with “/imagine”. Here is an example prompt I’ve used to create our own AI character (inspired by this TikTok video):
/imagine an astronaut on Mars, cinematic lighting, looking into the camera, detailed face, by art station, unreal engine q2
While writing a prompt, specify what you want clearly, try to visually define the objects, and be really descriptive, but do not overwhelm the system with too many details, and feel free to mention an artist in which style you would prefer your photo. Here you read additional tips for more efficient Midjourney text prompts.
If you would like to create an AI character that looks like you or a character you like, you can upload a photograph to Midjourney by following these steps:
Click on “+” and choose “Upload a file”. In this step, I’ve used a photograph of our founder and CEO Jero as a source image:
I’ve just removed the background from the image to make sure that AI doesn’t gather anything else from the pic. You can use Remove.bg for this step.
Once your file was uploaded, right-click on the photograph, and choose “Copy image address”
Create your prompt by inserting “/imagine”, your copied link, and text:
/imagine https://s.mj.run/a9XltgjMutY an astronaut on Mars, cinematic lighting, looking into the camera, detailed face, by art station, unreal engine q2
Once you have submitted your prompt, you should wait a bit until Midjourney creates your character. Also, as you are generating your image in a public channel, it might be confusing at first since you might not notice your picture - just scroll down or up a bit and your picture should be there.
If you would like to keep your prompts private, you will need to purchase the Private Visibility option add-on which will allow you to communicate directly with the Midjourney bot.
This is what Midjourney created for us on the first try:
Now that you have your picture generated, there are a couple of ways in which you can edit your AI character:
Making variations (marked as V1, V2, V3, V4 under the picture) - use this option to create different variations of a single image, you can pick one of the 4 generated images and create new ones
Upscaling (marked as U1, U2, U3, U4 under the picture) - use this option to upscale a picture you like the most for better image quality; once you scale the one you like, you can create additional iterations of it
Rerolling - this option generates a new set of 4 primary pictures, use it if you are not satisfied with the results at all
Additionally, you can make variations for each photo that you upscale by entering an additional text prompt after clicking on “Make variations” (for example if you wish to add any element to your character).
I’ve decided to upscale two of the generated versions before making a decision on which one to go forward with.
The winning photo was the one on the right as this character was looking straight into the camera, which was one of the most important parameters of creating this AI character. Not to mention that this character already looked like young Jero so double-win! 😃
If for any reason, you cannot find your generated character based on your prompt, go to this website, log in with your Discord server and you should be able to see all generated pictures. Here is an example of a prompt with all the information you can see when you log in:
Although I’ve created the voiceover and the text used in the post on my own, if you want to spare some time, you can ask ChatGPT to create the voiceover or a complete social media post for you.
An example of an entire Linkedin post generated by ChatGPT:
An example of a voiceover generated by ChatGPT:
Compare ChatCPT prompts with the text and voiceover written manually in the post and let us know which one is better 😄
There are many text-to-voice software available online and those that are way better, often are quite expensive (such as Well Said Labs). Since I’ve decided to create a voiceover really quickly, I’ve chosen to use Play.ht to generate it. If you would like to generate a more natural-sounding voice, I would strongly recommend giving WellSaid Labs a try.
To get started with Play.ht, create an account, insert your text voiceover and spend some time finding a proper voice that fits well with your created AI character. If you do not like how voices sound on Play.ht, feel free to look for any other text-to-voiceover tool online.
Creating voiceover with Play.ht tool:
Once you have generated your voiceover, click on “download” and save your preferred sample. For more customization, I recommend saving each clip separately instead of using only 1 clip as your sound will sound better. You can always merge audio tracks later on through a tool such as CapCut.
Once you have your character photo and your voiceover ready, it’s time to animate your character in a character generator tool! Head over to D-ID studio, register, and click on “Create video.”
Here you will notice a bunch of templated presenters you could use for your video, but if you have created your own, personalized character, add it by clicking on “+ Add”.
Pick your character photo and on the right choose “Audio” and click on “Upload your own voice” to add your own sound.
D-ID also has some predefined sounds you can use, but from my experience, the ones created through other tools tend to sound more natural, but give it a try and select the one you like better.
Now that you have selected your character photo and uploaded your voice, click on “Generate video” in the top right corner:
After a couple of seconds, your video is ready! If you feel like you can adjust any of the elements (picture, text, voiceover), make iterations.
If you are already satisfied with the result, you can either download the generated video or share the link on social media directly.
My recommendation would be to download the video and make the final polishing touches in CapCut or any other video editing software.
Another optional step would be adding subtitles to the video that are in line with the voiceover.
This is really convenient if you plan to post your video on a social media network that turns off the sound automatically when playing a video to grab user attention.
I’ve used a tool called Veed to automatically generate subtitles. Not to worry, if the tool doesn’t generate a 100% accurate subtitle, you can change it on your own.
Once you have created your AI character video with a voiceover (and subtitles), you are ready for the next step, and that is editing in video editor software.
I’ve used CapCut to create a different video format and edit the video settings. I’ve also added background music for a cooler atmosphere. Here you can check out a full video of the created AI character. And that should be it - your AI character is ready to conquer the Internet! You can always iterate and upgrade your AI character over time. Happy building!
Like what you’ve read? Check out our open positions and join the Martian team!See openings