FLUX 1.1 Deep Dive (NEW Frontier of Hyper-Realistic AI Photography)

Рет қаралды 24,552

Күн бұрын

Пікірлер: 48

@cyberjungle 3 ай бұрын

Prompt Tip: Try including "IMG_1025.HEIC" in your prompts to get very realistic images with FLUX1.1. Example prompt: IMG_0134.HEIC: a dog

@LouisGedo 3 ай бұрын

Hi 👋

@waitaminutepause 3 ай бұрын

It's not man is carrying...it's just man carrying

@cyberjungle 3 ай бұрын

@@LouisGedo hello ✨

@yoagcur 3 ай бұрын

@@waitaminutepause That does not make much difference. With correct English, It still ended up with half horse-half baggage. None of the models nailed that prompt

@Retro_Movie_Rewind 3 ай бұрын

Great video, thanks! Even as someone who uses AI tools every day, it's so hard to keep up with what's new. So I really appreciate your excellent guides.

@cyberjungle 3 ай бұрын

💚

@sachucoinage3533 3 ай бұрын

I understand you are trying to be mild while talking about midjourney instead of being rude but it is a fact that midjourney is loosing its charm. They are acting like nokia of the past thinking they rule the market. They are wasting time with philosophies and rooms and what not while the competitors are busy improving there product. I have been a long time supporter of midjourney but they have started to irritate me with only regular office hour promises and a few features here and there. Only thing saving them right now is there superior aesthetics, which i don't know for how long.

@Souls_from_darkness_to_Light 3 ай бұрын

They’re woke and support SDMY big time. If they’re not careful with that they’ll be history

@scitechaiutopia 3 ай бұрын

Midjourney is been all about money and subscriptions, their sole purpose of existence. So far I found Flux landscape images to be far better, Flux is also widely available.

@yoagcur 3 ай бұрын

@@Souls_from_darkness_to_Light They are woke in asking for suggestions and then just ignoring what their community asked for. Still produce the most stunning images, though

@atlanteum 3 ай бұрын

Thank You!! MJ seems to get WORSE with every iteration - not in its ability to produce beautiful imagery, but to follow the simplest prompts. I would trade speed and cost for ACCURACY any day if Flux or other models can provide it. And don't even get me started on the MJ exclusions, rules and moderators...

@barackuse 3 ай бұрын

I can't imagine where this technology will be in three years.

@Naejbert 3 ай бұрын

A challenge will be to make sure the models don’t learn from all of erroneous images generated today…

@ThierryPianoSolo 3 ай бұрын

@@Naejbert Ha ha ha, good point! ;)

@prod.bykbrd5723 3 ай бұрын

You have always been my greatest teacher since day one. Thank you❤

@cyberjungle 3 ай бұрын

💚

@neurojitsu 3 ай бұрын

Great overview thanks. I'm really hoping MidJourney bring out story telling mode soon, as it would fit my workflow needs. If it's good, it will really differentiate them too.

@cyberjungle 3 ай бұрын

I hope so too!

@MichaelWinterberg 3 ай бұрын

The era of 'versus' videos is slowly coming to an end, isn’t it? The differences are getting smaller and smaller and smaller and smaller

@cyberjungle 3 ай бұрын

Yes 👍🏼

@nesrinunur2537 3 ай бұрын

This is an excellent video.❤

@holdthetruthhostage 3 ай бұрын

Hopefully you cover text generation as well & Midjourney was to lazy its over for them

@reasonsreasonably 3 ай бұрын

Might have been better to have a double blind test by using assistant to gather the pictures for you to then gauge to remove your bias.

@cyberjungle 3 ай бұрын

100% 👍🏼 I will do this in my next comparison. Great idea

@A.M.-artvideo 3 ай бұрын

🔥

@holdthetruthhostage 3 ай бұрын

Hopefully you cover text generation as well

@gnoel5722 3 ай бұрын

Midjourney needs to get their priority straight. They are getting owned left and right and still thinks they are the king. The fact that 80-90% of my generations in Midjourney still has mistakes/ugliness with fingers and eyes OR cant get a text prompt correct is baffling. The only thing they have going for them is their character consistency but even then, like you said, you can train a Lora and will get even better results. They should stop their wokeness nonsence, unblock the NSFW and get a decent webUI / edit tools if they want to stay relevant.

@puja1985 3 ай бұрын

These days, we're so busy comparing tools that we forget to create. Let's focus on producing and nurturing our shared human potential

@stevemuzak8526 3 ай бұрын

People want to consume. Not create.

@holdthetruthhostage 3 ай бұрын

Where is Indeo Gram

@sandrovaro 3 ай бұрын

It’s free??

@kuko61 3 ай бұрын

For me the most important thing is the atmosphere and soul that the images generated in MJ have with the right prompt. Flux has a very good output, the consistency and image quality is excellent, but the images are too polished, clean and colourful for me. Achieving the atmosphere or photographic effects that I am able to create in MJ is often very difficult or impossible to achieve in Flux. One thing to disagree with in these comparisons is comparing the Mystic or Mystic V2 to other models. Mystic is not a model in itself. Mystic 1 was Flux + Magnific creative upscaling, I assume Mystic 2 is similar. If Mystic comparisons with other models were to be fair, the Flux or MJ outputs would need to be upscaled with the Leonardo Upscaler or Magnific. Only with such image enhancement and resolution increase would they be fully comparable to Mystic. And to the skin in MJ. Even in 6.1 I can achieve a very relistic skin. Not always, but in most cases. And of course the result can be improved, not only with sref, but also with training and personalization. Of course you can't compare it with Mystic, but as I wrote, with Mystic it's the creeative upscaling and high resolution that make the biggest difference.

@amykpop1 3 ай бұрын

Ideogram v2 beats them all imo, especially when it comes to prompt understanding.

@cyberjungle 3 ай бұрын

I will make a deep dive for Ideogram V2. On my list 👍🏻

@kuko61 3 ай бұрын

Yes. But in terms of image quality, the Ideogram lags behind the other models mentioned here. I would like to see a model that has the understanding of the prompts that the Ideogram has and the image quality of the MJ or Flux.