생각해 보면, 올해초만 해도 Kaiber의 영상 보면서 AI영상생성 대단하다고 느겼는데 , 지금 보면 상당히 유치했죠. 1년도 안된 사이에 이렇게 까지나 발전했어요. 영상에서 모든 이미지나 동영상 생성이 모두 너무나 뛰어난데 그 중에서 더 나은것을 고르고 있을 정도가 되었네요. 앞으로 1년은 또 어떨까요? 정말 빠르게 발전하네요. 늘 좋은 컨텐츠 감사합니다. :) When I think about it, at the beginning of this year, I was watching Kaiber’s videos and thought AI-generated videos were amazing. But now, they seem quite rudimentary. In less than a year, we've seen such immense progress. Every image and video generation tool now performs at such a high level that we’re even able to compare them and choose the best ones. What will the next year bring? The pace of advancement is truly incredible. Thank you for consistently creating great content. :)
Problem with Runway is it makes everything G-rated. I'm working on an AI film (that is now 30 min long...) and there's a scene where a woman is being chased by a mob of angry women in the Middle Ages. Well, Kling knocked it out of the park. Vidu also was insanely good for the action, less so the resolution. But Runway? It made all the women jogging together. And I'm talking in multiple attempts with different prompts. All were based on the same source images, as well. And for another scene I just needed a waitress screaming. Runway made her look off in the distance as if in a trance. And when I needed someone to kiss someone else on the cheek, Runway refused to make the physical contact. So, great for doing Disney stuff for the kiddies. But if it's PG, nah.
@@curiousrefugeI see what you mean. There are major flaws with maligned prompt interpretation as it relates to undefined terms without full legal definition and context, misclassifications, and opaque moderation processes that unfairly punishes users for unwanted content never even received. Current TOS hurts creativity and artistic expression. I will drop them a note with some observations and ideas.
The problem with AI images is that they always make the background blurry as if it was taken with an F2 depth of field lens. And it is not possible to make a smaller aperture such as F8. In general, I think that everything about creating pictures and movies in AI is still ahead of its time, it reminds me of the Internet in the mid-90s, which took about 4 years from the time it was popular until it more or less stabilized in terms of performance.
Try describing detail in both foreground and background in you prompt. Also find a photographer who specialises in these sharp images. Ask google gemini or chat gtp for more words describing the "no depth of field" result you want.
I get 25% deep focus shots with this style of prompt: A hyper-realistic, sharp, detailed, ultra-clear, cinematic wide shot with deep focus closeup of a man in futuristic clothing standing with a futuristic cityscape displaying a clearly crisp clear focussed massive skyscraper in the background, deep focus, everything in focus, Leica M
@@curiousrefuge My intention is that by default AI generators produce F2 depth of field if I don't write a long prompt. Note that most of the photos in your video also have a blurred background, probably the AI engines learned with professional photos with a blurred background or with amateur photos with a clear background F8 but poor quality of someone who doesn't know how to take pictures at all. I don't think anyone bothered to teach AI engines professional photos but different lenses and different settings of a professional camera. The second thing that bothers me about AI is that every time I enter a beautiful woman or a model in the prompt, it gives me women full of makeup and huge lips from several plastic injections. Never a naturally beautiful woman.
I was at the West Palm Beach meeting for only 20 mins. I got there late, it was almost done and I didn't want to bother anyone there. Hopefully they can do another one in Miami or Fort Lauderdale area! :)
Did you draw the conclusion that Flux Pro Ultra was the best with only realistic images? How does it fare with other styles, or creating things that doesn't exists, not things that you can actually take a photo of?
Great video! I’ve subscribed and really enjoyed your take on this topic. I’m currently working on videos covering the latest AI news, so it’s awesome to see the community growing with such valuable content. Keep up the fantastic work!
Even with image to video, Kling seems to maintain the look of the character's face...whereas Runway seems to have a library of video faces it uses that can only roughly approximate the image.
Luma started great but it has been falling behind the other platforms lately. Hopefully, they're working on some special to be released to the public for Xmas ?
There was one terrific image generator you left out of the comparison: Magnific's latest version. It beat everyone else by quite a bit. It's realism is superb, especially of faces. I haven't tried Flux Pro Ultra but I will. As always, great information and great presentation.
my favorite image generator missing from the comparison but .. then it still feels like my favorite indie band. That at some point everyone will know about.
@@curiousrefuge ideogram. the interactions between humans and prompt adherence is insane and has been since day 1 - but v2 is still my fav. That mixed with minimax is hands down the most impressive combo imo
I hear you. I still like midjourney for development and having all the tools in one place, but for simply the best realism and deep character consistency (training models) it’s not the best.
the game generation part is BS. It doesn't actually create a map , it creates scenes and fails even in that. We're not there yet. The game wouldn't make sense as the map would be fluid . Imagine you want to go back to where you were but that place doesnt exist any more. It would feel like dreams do
CC : I like your vids... but cut out the irrelevant fluff/extending the length of you videos. Volume for the sake of volume is a negative. Gotta skip through your vids to get to what matters.
I think there is some randomness in comparing the images. I am a BEEKEEPER and I get amazing new photos out of Midjourney. I know what a bee looks like! (-:
I have to correct you again. Disney built this unit to investigate and go after models stealing their patents. Pixar style? Does it sound familiar? Well, it is copyrighted and trademarked. they’re not seeking to use AI anytime soon in active productions. And this comes from someone from the team. Say no more.
Interactive entertainment is also a huge time suck. A blockbuster film could take two years from development to final product. A good video game might take longer...but you can get a lot more play out of it. Let's take Baldur's gate. You can easily clock 100+ hours and replay it multiple times with different results.
If you sell AI video to a client you must disclose that it is designated as Public Domain by the US Copyright office and anyone can use it. They can’t own it. If you lead them to believe that they own it you could be held liable in a court of law!
Ah, this isn't quite accurate. It’s true that the U.S. Copyright Office generally doesn’t grant copyright to art created solely by AI without human input, as they require a "human author" to be eligible for copyright. However, if somebody has creative involvement in the AI-generated work ( like by making selections, edits, or arrangements ect,) that human-authored portion can indeed be copyrighted and owned.
Yeah we had IP lawyers in our office hours and they affirm what Brock said. If the project had some human intervention and direction, which virtually any AI project in the world does, (previs, prompt guidance, editing, sound mixing, etc.) you can apply for copyright. That doesn’t mean that 100% of the time it will be granted. For example if you did minimal changes (like light generative fill) to an image and you’re looking for a copyright on that image there might be an issue, but for a video it isn’t so black and white. The most comparable example would be a documentary that has public domain assets in it. The assets may be non copyrightable, but the edit and final doc is because a human compiled it. And that’s not to mention lora training and custom models.
@@brockhampton3078 That is highly debatable, and the answer will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work. This is necessarily a case-by-case inquiry. Nevertheless, the individual shots out of context are still considered public domain and free to use unless combined with significant human creation of elements within it. If a work's traditional elements of authorship were produced by a machine, the work lacks human authorship and the Office will not register it. www.federalregister.gov/d/2023-05321/p-44
Corridor Crew on youtube did a spoof on George Lucas and his many tweaks and fixes he did on star wars. It made the prequels look hilarious. Thats why I stick to the original trilogy. Jar Jar Binks still sux. Lol 🤣
@ Don't be coy. "Better than MJ" "MJ is 3rd?" and a number of your comparisons infer that some new program could spell the end of MJ. Address the issue... ie. you don't have the first clue how to achieve the best results from MJ, so you're wasting everyone's time making comparisons until you do. Can't take you seriously. About to select " Don't recommend this channel." That means bye-bye's... forever. Tip: If you're going to make comparisons between products, compare apples with apples. Take a little time to understand WTF you're talking about before put two images side by side for criticism. Be fair, not dodgy. You're welcome.
Thank you for sharing my work. I am a fan of your project)
Your work????
Thank you very much!
Why is AI Me so ANGRY?!!!!! I'm just happy he takes all that aggression to the gym to get JACKED!
Dial back the roids bruh
Dial back on the roids.
ㅋㅋㅋ 화난게 아니라 그냥 진지해 보여요. ㅋㅋㅋ 오셨네요. ㅋㅋㅋ
All those roids you've been taken for all these years!!!! JK we
I'm happy with Runway... but this Ultra Model seems amazing! I also love the userinterface for choosing aspect ratios, etc.
So cool, right?!
생각해 보면, 올해초만 해도 Kaiber의 영상 보면서 AI영상생성 대단하다고 느겼는데 , 지금 보면 상당히 유치했죠. 1년도 안된 사이에 이렇게 까지나 발전했어요. 영상에서 모든 이미지나 동영상 생성이 모두 너무나 뛰어난데 그 중에서 더 나은것을 고르고 있을 정도가 되었네요. 앞으로 1년은 또 어떨까요? 정말 빠르게 발전하네요. 늘 좋은 컨텐츠 감사합니다. :) When I think about it, at the beginning of this year, I was watching Kaiber’s videos and thought AI-generated videos were amazing. But now, they seem quite rudimentary. In less than a year, we've seen such immense progress. Every image and video generation tool now performs at such a high level that we’re even able to compare them and choose the best ones. What will the next year bring? The pace of advancement is truly incredible. Thank you for consistently creating great content. :)
Crazy how fast the entire ecosystem is moving!
Problem with Runway is it makes everything G-rated. I'm working on an AI film (that is now 30 min long...) and there's a scene where a woman is being chased by a mob of angry women in the Middle Ages. Well, Kling knocked it out of the park. Vidu also was insanely good for the action, less so the resolution. But Runway? It made all the women jogging together. And I'm talking in multiple attempts with different prompts. All were based on the same source images, as well. And for another scene I just needed a waitress screaming. Runway made her look off in the distance as if in a trance. And when I needed someone to kiss someone else on the cheek, Runway refused to make the physical contact. So, great for doing Disney stuff for the kiddies. But if it's PG, nah.
True, we wish they could tone down the settings a bit to allow for more material!
@UnknownFear-sr9un Kling, Vidu, and Minimax (Hailuo) would work a lot better.
@@curiousrefugeI see what you mean. There are major flaws with maligned prompt interpretation as it relates to undefined terms without full legal definition and context, misclassifications, and opaque moderation processes that unfairly punishes users for unwanted content never even received. Current TOS hurts creativity and artistic expression. I will drop them a note with some observations and ideas.
Yeah this is a deal breaker for me aswell
The problem with AI images is that they always make the background blurry as if it was taken with an F2 depth of field lens. And it is not possible to make a smaller aperture such as F8. In general, I think that everything about creating pictures and movies in AI is still ahead of its time, it reminds me of the Internet in the mid-90s, which took about 4 years from the time it was popular until it more or less stabilized in terms of performance.
Try describing detail in both foreground and background in you prompt.
Also find a photographer who specialises in these sharp images.
Ask google gemini or chat gtp for more words describing the "no depth of field" result you want.
I get 25% deep focus shots with this style of prompt: A hyper-realistic, sharp, detailed, ultra-clear, cinematic wide shot with deep focus closeup of a man in futuristic clothing standing with a futuristic cityscape displaying a clearly crisp clear focussed massive skyscraper in the background, deep focus, everything in focus, Leica M
You can definitely modify your prompt. what is your prompt?
@@curiousrefuge My intention is that by default AI generators produce F2 depth of field if I don't write a long prompt. Note that most of the photos in your video also have a blurred background, probably the AI engines learned with professional photos with a blurred background or with amateur photos with a clear background F8 but poor quality of someone who doesn't know how to take pictures at all. I don't think anyone bothered to teach AI engines professional photos but different lenses and different settings of a professional camera.
The second thing that bothers me about AI is that every time I enter a beautiful woman or a model in the prompt, it gives me women full of makeup and huge lips from several plastic injections. Never a naturally beautiful woman.
I was at the West Palm Beach meeting for only 20 mins. I got there late, it was almost done and I didn't want to bother anyone there. Hopefully they can do another one in Miami or Fort Lauderdale area! :)
There will be another soon! :)
Palm Beach? Dude, I'm in Savannah! I had no idea you were so close! I need to know about the next meetup!
I ran the West Palm Beach Meetup, it was a great first turnout too.
@@classicvwbugs Please keep me informed of any other ones! I will definitely be there!
@aaronhhill December 5th is the next one
Mama Bella's in West Palm Beach
@@classicvwbugs Awesome! Looking forward to it!
Love your content! Would have came to the West Palm meet up if I knew about it.
Always the best AI news source! Thank you!
Is it possible to create a character using text to image, that has no background and can be superimposed onto another background I create separately?
You can create it and then easily remove the background iwth a variety of tools.
@@curiousrefuge Thank you!
Did you draw the conclusion that Flux Pro Ultra was the best with only realistic images? How does it fare with other styles, or creating things that doesn't exists, not things that you can actually take a photo of?
It may be best - but remember so much of this is also about having control!
Who picks the films of the week? Are films submitted?
They are curated from everywhere and chosen by our staff :)
@@curiousrefuge I got one if you would like to take a look? I'm the guy running the WPB meetup.
Great video! I’ve subscribed and really enjoyed your take on this topic. I’m currently working on videos covering the latest AI news, so it’s awesome to see the community growing with such valuable content. Keep up the fantastic work!
Awesome! Thank you!
Where can we watch those THE LAST MORTAL and other videos?
Check our gallery : )
@@curiousrefuge went through channel and website but couldnt find it.
The Ultra version of Flux, does give some awesome output from people to architecture.
Flux is awesome!
Even with image to video, Kling seems to maintain the look of the character's face...whereas Runway seems to have a library of video faces it uses that can only roughly approximate the image.
Kling is quite powerful!
Luma started great but it has been falling behind the other platforms lately. Hopefully, they're working on some special to be released to the public for Xmas ?
We'll see!
Great job bro.
I appreciate it
There was one terrific image generator you left out of the comparison: Magnific's latest version. It beat everyone else by quite a bit. It's realism is superb, especially of faces. I haven't tried Flux Pro Ultra but I will. As always, great information and great presentation.
Thanks for sharing!
my favorite image generator missing from the comparison but .. then it still feels like my favorite indie band. That at some point everyone will know about.
What is it?
@@curiousrefuge ideogram. the interactions between humans and prompt adherence is insane and has been since day 1 - but v2 is still my fav. That mixed with minimax is hands down the most impressive combo imo
you're not even using flux pro. on the bottom left, you have to select the right model in krea
thanks!
In the past I would create macros to complete tasks for me on the computer, AI is becoming a macro for humanity.
Interesting perspective!
12:00 future VR/AR 'holodeck'
We can't wait :)
Midjourney remains the best stylistically and artistically.
yeah, the character consistency and different remixing options is the biggest plus for me in midjourney.
I hear you. I still like midjourney for development and having all the tools in one place, but for simply the best realism and deep character consistency (training models) it’s not the best.
Mar-say, Caleb!
If you guys don't hurry up and get to Austin, I'm going to lose my s***! Thank you and love you guys!
We can try to get a AI Meetup in Austin soon :)
Thanks 🙏
You’re welcome 😊
Doesn anybody have a working link to flux pro ultra?
We'll see what we can find!
the game generation part is BS. It doesn't actually create a map , it creates scenes and fails even in that. We're not there yet. The game wouldn't make sense as the map would be fluid . Imagine you want to go back to where you were but that place doesnt exist any more. It would feel like dreams do
It's very...very very early in that tech!
CC : I like your vids... but cut out the irrelevant fluff/extending the length of you videos. Volume for the sake of volume is a negative. Gotta skip through your vids to get to what matters.
We appreciate the feedback!
we bouta have a holodeck
hopefully soon!
I think there is some randomness in comparing the images. I am a BEEKEEPER and I get amazing new photos out of Midjourney. I know what a bee looks like! (-:
I was wondering about this. Midjourney always gives 4 results, but he only shared one. I am getting terrible hands very reliably out of MJ these days.
We'll work on showing more next time :)
It appears access to Flux Pro Ultra requires a paid version of Krea - is that right?
That sounds right
@curiousrefuge I just signed up for NightCafe pro instead
@curiousrefuge in accessing thru NightCafe now
Cant beat Midjourney for stylistic images, theyre no 1.
MJ has instant lora in style ref, and you can control the weights of each style as well
MJ is very powerful in terms of control!
20:06 20:06 attention professional animators: get your heads out of the sand before our industry is over run by tech bros like this!!!
Thanks for watching!
I have to correct you again. Disney built this unit to investigate and go after models stealing their patents. Pixar style? Does it sound familiar? Well, it is copyrighted and trademarked. they’re not seeking to use AI anytime soon in active productions. And this comes from someone from the team. Say no more.
I'm afraid your wrong on this one.
Interactive entertainment is also a huge time suck. A blockbuster film could take two years from development to final product. A good video game might take longer...but you can get a lot more play out of it. Let's take Baldur's gate. You can easily clock 100+ hours and replay it multiple times with different results.
Very true!
No AI compares to Midjourney in imagination!
MJ is pretty powerful!
"I love unicorns so so" 🦄🤣
That part about the super speed zoom around the guy was funny
hahah glad you enjoyed that!
If you sell AI video to a client you must disclose that it is designated as Public Domain by the US Copyright office and anyone can use it. They can’t own it. If you lead them to believe that they own it you could be held liable in a court of law!
Ah, this isn't quite accurate. It’s true that the U.S. Copyright Office generally doesn’t grant copyright to art created solely by AI without human input, as they require a "human author" to be eligible for copyright. However, if somebody has creative involvement in the AI-generated work ( like by making selections, edits, or arrangements ect,) that human-authored portion can indeed be copyrighted and owned.
Yeah we had IP lawyers in our office hours and they affirm what Brock said. If the project had some human intervention and direction, which virtually any AI project in the world does, (previs, prompt guidance, editing, sound mixing, etc.) you can apply for copyright. That doesn’t mean that 100% of the time it will be granted. For example if you did minimal changes (like light generative fill) to an image and you’re looking for a copyright on that image there might be an issue, but for a video it isn’t so black and white. The most comparable example would be a documentary that has public domain assets in it. The assets may be non copyrightable, but the edit and final doc is because a human compiled it. And that’s not to mention lora training and custom models.
@@brockhampton3078 That is highly debatable, and the answer will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work. This is necessarily a case-by-case inquiry.
Nevertheless, the individual shots out of context are still considered public domain and free to use unless combined with significant human creation of elements within it.
If a work's traditional elements of authorship were produced by a machine, the work lacks human authorship and the Office will not register it.
www.federalregister.gov/d/2023-05321/p-44
How many years do you think it will be until Ai movies can be made without having to cherry pick videos? Ai photos are almost there..
Less than 5!
Perplexity is a scab? LOL
a scab?
Unity 3D, Unreal should be worry.
Interesting point!
Corridor Crew on youtube did a spoof on George Lucas and his many tweaks and fixes he did on star wars.
It made the prequels look hilarious. Thats why I stick to the original trilogy.
Jar Jar Binks still sux. Lol
🤣
haha we love corridor crew!
the minecraft AI thing is not related to google dont clickbait
What do you mean ?
@@curiousrefuge just that oasis is not made by google
Oh, another AI that "beats" MJ... and yet they never do...
We appreciate you watching!
@@curiousrefuge I appreciate your videos. Just hate the clickbaits, but that's how internet works...
can you give me your course for free i will pay when its make money for me in youtube
Nope.
Every time I see this title I get disappointed
Do you not feel like it has better quality?
✝️ Subbed/Liked! Great VID! God bless in ‘24! ✝️
Appreciate you!
It's bokeh, not bokah.
A reminder for those people who don't understand why nobody ever invites them to parties. This is why.
🚔🚨
Sorry!
More Midjourney bashing by someone who doesn't have the first clue about how to prompt in MJ to achieve the best results. 👎
We love MJ, what do you mean bash?
@ Don't be coy. "Better than MJ" "MJ is 3rd?" and a number of your comparisons infer that some new program could spell the end of MJ. Address the issue... ie. you don't have the first clue how to achieve the best results from MJ, so you're wasting everyone's time making comparisons until you do. Can't take you seriously. About to select " Don't recommend this channel." That means bye-bye's... forever. Tip: If you're going to make comparisons between products, compare apples with apples. Take a little time to understand WTF you're talking about before put two images side by side for criticism. Be fair, not dodgy. You're welcome.