Stable diffusion VS Midjourney: All you need to know
Vložit
- čas přidán 31. 05. 2024
- AI art is one of the hottest topics when it comes to AI discussion in general, and there’s a question to be asked here. Is it possible to get a high-level AI image generation for free, or is the future exclusively behind paid services?
I’ve decided to compare two of the best examples in each category: Stable Diffusion vs Midjourney to find the answer.
-----------------------------------------------
⚖️Brief Comparison of Stable Diffusion vs Midjourney:
-----------------------------------------------
Currently, Stable Diffusion AI is an open-source text-to-image generator, freely available to anyone. It supports thousands of custom Stable Diffusion models, each tailored to a specific style, provides an extremely flexible customization model, and has a dedicated community that expands the possibilities daily. However, it’s also harder to run for an inexperienced user, and it requires quite a bit of learning to get a hang on.
Midjourney AI image generator is different in almost every aspect. It’s not open source, and using Midjourney is only possible with a heftily priced subscription. Comparing Stable Diffusion vs Midjourney, the latter is less customizable, and only has a couple of models, but they are extensive, and results created by Midjourney are of very high quality. It’s also more beginner-friendly, to the point where all you need to use Midjourney is a Discord account.
-----------------------------------------------
🤖 AI training:
-----------------------------------------------
Stable Diffusion is very straightforward in its approach: it learns how to generate images by destroying them. The idea here is to add layers of noise over and over again until there’s barely anything left of the original image, just so that AI will attempt to reverse the process and recreate the original from just a few scraps of data.
Midjourney is closed-source, of course, so all we can do is make an educated guess. Such a guess would suggest that Midjounrney combines the same approach of Stable Diffusion, plus a large language model, or LLM.
-----------------------------------------------
📸Models:
-----------------------------------------------
Starting with Stable Diffusion, there are many versions of the default model by the same name. Stable Diffusion is trained on a wide subset, so it does a good job in almost every style, though it’s not as detailed or nuanced as Midjourney.
Midjourney is not that customizable, at all, and relies on a single, constantly updated model. However, due to the more in-depth training, Midjourney has images of better quality, which come much closer to the prompt than Stable Diffusion.
-----------------------------------------------
🧑⚖️Can AI work be copyrighted?
-----------------------------------------------
In reality, it depends. As of August 2023, AI-generated art can’t be copyrighted in the US, since the copyright law only protects works created by human beings. In a recent case, a US court ruled that a work of art created by AI without any human input cannot be copyrighted because it lacks human authorship.
But, you might have noticed a loophole here. “Without any human input”. That’s right, if a human artist uses AI to generate images and then modifies or arranges those images creatively, the resulting work may be subjected to copyright as an original work of art by a human artist.
-----------------------------------------------
00:00 Intro
00:26 Brief Comparison of Stable Diffusion vs Midjourney
01:58 How does Stable Diffusion AI learning work?
03:00 How does Midjourney’s AI learn?
04:50 What models can Stable Diffusion and Midjourney offer?
06:30 Can AI work be copyrighted?
07:30 Final thoughts
-----------------------------------------------
#imagegeneration #aigenerated #midjourney #cooltechzone #ctz
Affiliate Disclaimer:
Some of the links on this channel may be affiliate links, which means we may earn a commission if you click on them and make a purchase. However, we only recommend products and services that we believe in, and our recommendations are always based on our personal experience and research. - Věda a technologie
Did I look cool in an Iron Man suit? 😎 Let's discuss AI in the comments!
Stable Diffusion XL has gotten way better in terms of interpretation of your prompts. It's very similar to MJ now, where you just vaguely describe what you want and don't need any negative prompts, and you still get what you want very consistently.
Stable diffusion is controllable to a large extent, especially if you know what you’re doing. Midjourney isn’t remotely controllable. It just produces pretty pictures, which are more-or-less randomly generated. So for any professional use Midjourney is purely a visualising tool which cannot be used usefully.
Using SD. It is technical but I can go deep with my imagination. You will run into problems but it is always fun to find a work around. If you just want a nice quick image go with MidJourney.
Yo, you have badass playlists 🔥
Supporting you as your beginner on CZcams in terms of numbers, but not in the valuable content you are providing in a very intuitive form.
Thank you
Thank you, that was a good comparison. I am currently using Midjourney but I want to try stable diffusion.
Underrated content, surprised you have such low subs. You've definitely earned one from me, thank you for the quick and insightful video :)
Much appreciated!
I've tried Midjourney, but had right away a love/hate relationship with it because of its commercial model. I felt in love at first sight with Stable's open source model, and although way more technical indeed, and less impressive when it comes to graphics (although this improves A LOT by learning how it works) I haven't felt anything really bothering me. To me it felt pretty much the same as discovering Linux and getting rid of Windows.
Only few days after having started to play with it, it pushed me to learn how to create my own models, and even getting into a huge project to help me and others to realize stuffs that were impossible, yet. It can't be emphasized enough how deep the rabbit hole of personal and collective contribution can go, nor how exponentially efficient Stable Diffusion will become.
Bless! had a research project on exactly this topic.
Such an underrated channel, bravo
Thank you!
Awesome video, love your teaching style and production value. I’m looking at generating realistic people that can also be customized to fit an emotion or reaction and still look like the same person. In your experience, which program would give the best results?
I'd say Midjourney is a bit easier to use in general, I struggled a lot with keeping same face throughout the process and best results came from Insight plugin
I may be relatively conservative in my political views, but the older millennial in me demands the open source softwares conquers the market and beats out the evil corporation.
So because a company wants paying for a product they’ve taken skill, time and resources to develop they’re “evil”. Wow. Just wow. Self entitled much?
Great video ...loved it ... subscribed...keep up the aweosme work... So why can't we add the language comprehension model with the SD ? sorry for the newbeee question ?
Hey, thanks a lot! Could you explain on your question? Do you mean foreign languages or additional LLM training?
@@cooltechzone I don't mean foreign languages ... I mean why can't we train stable diffusion with better or more LLM (English) to address it's inferior language comprehension issue ?
First of all SD primary function is diffusion, not language comprehension, SD was trained differently than other AIs', so training it with an LLM could be computationally expensive and resource-intensive. And as far as I know it would require a lot of complex modifications to the whole SD system.
@@cooltechzone thanks a lot ..keep rocking bro
1:16 for $10 you don't get Relax time. After the fast time is consumed, that's about it. I don't understand why this misinformation is so widespread.
great video. this is just what I need
GREAT VIDEO! THANK YOU!
I used LeonardoAI, and for me it’s the third placer for image generation…
Underrated channel
Thank you!
Only open source without any control or restrictions.
Hey nice artwork! Just want to ask that are you interested in our new ai generation product? Maybe I can send you beta version to you~
I will keep using SD to see my favorite female characters and celebrities in NSFW 🔞.... type images goddamn I love this technology.
midjourney is actually alien technology , this is why the tecnology is so hush hush
None. You glossed over the copyright issue, but it's way more complicated than that. It's theft. Simple as that. Even if an artist puts more work and paints over an AI generated artwork, that result still has work in them, that was created by others. The whole AI industry is also not profitable, and that even without compensating the artists and writers whose work is stolen and resused.
LAION on the other hand does not have the artwork and text themselves, because again, that would be against copyright. They just point at it.
And of course, there are lawsuits coming and running as well. And when you keep the costs of AI art in mind, than it is also not 'democratising' art. There is no expensive barrier to learn art. People simple don't want to put in the time to do it. That is not to say that the technology isn't amazing, but we are a long way from making it morally ok.
Spoken like a real ignorant. AI doesn't even store images but words. Are you going to argue that an image generated from random noise and image correction trying to match the noise with the words is a copyright issue?
@@Krullerized The MIT and other people doing studies, the US senate, lawmakers, copyright lawyers and specialists, they all see it differently. Before you call someone ignorant maybe begin with yourself. We're done here.
@@doppelkammertoaster Just goes to show what happens when people want to have opinions at all costs on topics they don't understand at all and ridicule themselves.
Image generator AIs don't have any image stored in their neural networks, only words and drawing instructions. They're just glorified image restoration programs trying to match random noise with words.
Maybe you should install one and see how wrong you were :)
Unless you just want to live in an echo chamber where you're right no matter what.
@@Krullerized True. But they still process the data the datasets point to. It has been proven by the MIT already. Before insulting other people maybe get your facts right. These stupid algorythms can to shit without all the data they steal.
you're really coping hard