Prompt engineering.

Hey! Important question!! I heard that Teespring does not promote your designs even you get the 10 sells. They have removed the trust score and you need to promote all you designs on your own??? That is what i heard, is it true?

Great video! I used a few different AI Art generators. I think the prompt should be more focused on the design style itself, not on what you are going to use it for. So don't use words like t-shirt design or poster etc., it takes that too literally. What works better, for example, is "A panda riding a bike in kawaii cartoon style" and in some generators, it helps also if you give more information like in the style of {artist name} example "van Gogh" or mood, environment, etc. In general, my conclusion is it takes a couple of tries to get something useful, but it can help with composition ideas or concepts. If lucky sometimes even useful parts can be cut out or can be used as templates/sketches to draw over with vectors. It can be helpful if you are stuck with ideas and get creative ideas. But mostly I am faster if I draw from scratch myself especially because I like my art in vectors. For generating text designs it is useless it mostly generates nonsense words.

Thanks for this intro! One thing is for sure, AI for designers is here to stay. Will be … interesting … to see where it all goes!

Couple of points:
1) I’ve heard that the new Deepfloyd software can handle text, but other than that, text is a known weak point for the ai art generators, and canva is likely using one of the older models
2) You usually need to put in hundreds of repetitions with good prompts and even in-painting (deleting a section of an image, and then regenerating that section with the same or a different prompt) to get to the image that you want. That said,
3) I think that currently AI art will be useful for base images that standard graphic design apps then alter into a final design, and a GODSEND for Lifestyle ads and stock photos. Also
4) AI art is a huge topic that is constantly developing. I recommend the channels aitrepreneur, Mattvidpro AI, and your best scuba kit if you’re interested in learning more.

I do currently my master in artificial intelligence. I think Canva has just fear to get hanged up by the competitors. Under the hood, i suppose, they use Stable Diffusion or Dall-E. Nothing impressive from Canva, im sure, just fear-driven.

❤️

Thanks Juna for this video. I've tried this feature at Canva and it really depends on the prompts you type. I have not tried to use it in POD.

Can I use this for print on demand product?

Yes, this service needs a PRO account.

use midjourney is better than canva

I tried it out and the even if the technology is new the images it creates are shockingly detailed and borderline scary

Thanks for the video, it was a great help with using Canva in general…

It certainly has it's strengths, but it doesn't comprehend T'-shirt design. I have had success with patterns, surface and seamless, but some work and some don't. It is definitely trial and error. My best have been some images of horses, cats, unicorns, mention specific colors mainly concept art works the best. Text is an epic failure. I think if you generate the graphics and then incorporate other elements and text, it could probably be useful. I would leave out the word T-shirt design and just be specific on the graphics you want to create. It doesn't understand styles of design, text, or the size of the space you are trying to fill. It has potential, but a long way to go.

Juna, I am not sure if you know who Hamish is, he has a channel here on YT. He just did a video about this, but he used Midjourney, however, he names a channel in his video, that showed him what prompts to use. Anyway, I don't want to name the channel, for two reasons, the first YT will block my comment, and the second is, there are several channels with the same name, so it's better to watch Hamish's video, he shows the name of the channel at the beginning of his video. I think he made the video yesterday, or the day before. It might be helpful for you to get better images. The prompts are key. Using simple prompts doesn't result in very good images. The more descriptive words you use, the better, and also add in art styles and names of artists. It is really about the prompts.

Oh and there is a channel I watched, about upscaling, if I can find the video, yesterday, I will tell you what it is. You can take the image into "bigjpg," then take it into Photoshop and you can really make it good. Bigjpg will upscale it 4X, for free, but if you want it bigger it is a paid subscription.

The video for upscale is called

The best way to upscale AI art for printing by Analog_dreams.

Hopefully, YT won't take my comment.

Canva text to image was an app for the free account, I am not sure if it still is, though.

Thanks for the video, Juna!

Hey Juno,
I always enjoy your videos. You should try Midjourney (I think it's the best I've tried so far), Dall-E, and Creative Fabrica Spark. Would like to see videos with you testing these AI tools and maybe other ones I don't know about.
Thank you for your content.

Very informative video! Thanks, Juna, I think text-to-image generators can be used to generate unique graphics, that no one else used on their t-shirt. The rest is, adding the text, and generating the design! Maybe you can try Midjourney in your other videos. It is a wonderful tool.

They all look like crap!

A very informative video
Thanks for sharing