62+ Astuces Fouillées pour Exploiter la Magie de la Génération d’Images avec ChatGPT ! (GRATUIT)
Bien sûr ! Pour vous aider à créer un résumé journalistique, il me faudrait le contenu de la transcription de la vidéo YouTube que vous souhaitez résumer. Pourriez-vous s’il vous plaît me fournir le texte ou les points clés que vous souhaitez inclure ?
Source : AI Samson | Date : 2025-04-17 14:26:21 | Durée : 00:33:19
Download the FREE prompts: https://delightfuldesign.eo.page/3m564
Try Virbo to create stunning AI B-roll in seconds:
https://bit.ly/3FF2od2
Thanks to Virbo for sponsoring this video.
nice nice
What? Figma to crop out images??? That is a WEB DESIGN tool. You would use something like Photoshop (or any free alternative) to do that, not Figma.
I love your videos Dude
Well, the segment where ChatGPT created comics featuring itself was really something 😀 Interesting
Great video!
The main issue is that you will NEVER get 100% what you want. And the more changes you ask, the more it will definitely not look like what you wanted. From iteration to iteration you'll loose the otiginal image or prompt completely so the only option is to start over. Keep in mind that you are paying for all the bad renditions as well with money and time (and nerves). This is why this is not a viable tool for any SERIOUS creative work. It's a fun novelty, yes, but nothing more.
Product photos with products ever so slightly changing from picture to picture? I would definitely not want that. Also, the AI is ruining the texts and logos on products.
These mockups are trash sadly. Zero typography, very little appeal and when you start to iterate them, they fall apart. This technology WILL be good at some point, but by then we won't be needing it anymore. This just gives YOU the impression that you have created something that works and has value, while in reality, you have not. The mockups are soullless and completely lack creativity.
It was rather satisfying to realize that what was sold by the thumbnail is not actually in the video before wasting 33 minutes of my life.
Ok, I'll admit it. You got me with that thumbnail! 😅
Hi, great video! Can you recommend an ai video generator that can generate product videos or lifestyle mockup video?
Well done! Plenty of workable ideas right there. Thanks
More hyperbole about the capabilities of AI. I knew there was a “masterclass” sales pitch coming 😂
We’ve been through so many hype cycles now (drop shipping & print on demand, Web3/crypto to name a few). There’s always someone on YT claiming to be an expert and shilling their course.
OpenAI’s rendering engine is technically impressive, but often unusable due to the ridiculous amount of so-called “ethical” restrictions.
Many of these limits seem completely disconnected from real-world ethics, making this feel more like Artificial Stupidity than Artificial Intelligence.
It’s a massively powerful system crippled by absurd over-cautiousness.
We seriously need to rethink what’s truly ethical… and what’s just performative control.
I wouldn't say that absolutely everything described in the video works as well as I would like. The saddest thing is that GPT can't take 100% of the image you gave it and reproduce it. I tried to transfer my prints to a T-shirt, or use my logo, but it doesn't take the attached images 100%, it looks at them and generates them based on what it saw, so it turns out to be a parody. Yes, it generates excellently from 0 if there are not very many small elements, but unfortunately it can't use the material given to it. This is even visible in the video where the author uploaded broken glass to it, it redrew it in its own way. The situation is even worse with the faces of the uploaded people. GPT is capable of reproducing 90% of reliable public figures: musicians, actors, politicians, even by searching the Internet it can find the Mayor of a small town, but it reproduces any uploaded faces by 40%, some features are similar, but otherwise a completely different person. As was agreed in the discussions, this is done to avoid deepfakes of real people and similar scams. In the best case, you can make similar caricatures, but you will not achieve a realistic photo as shown in the video with the shark, the face will never be yours. And even if you give it a link on the Internet to a publicly available photo of a person on the Internet, it will refuse to do it, because there was no agreement to do this, even if it is you yourself. They say that in the first 2 days when this model was released, it could perfectly make 100% copies of uploaded faces, now, alas. Otherwise, an amazing tool.
Also regarding creating a drawing with a transparent background, there are problems here too. It makes the background transparent yes… at first glance. But in fact, it has a huge number of garbage pixels all over the background, they are not noticeable to the eye in the picture itself, but if you use this image for serious needs, all this pixel garbage comes out, so this image will still need to be cleaned manually along the contour. Yes, this is better than cutting from scratch, but this needs to be mentioned. There is also a downside to this problem, the image itself also has pixels, but now they are transparent, alpha channels, i.e. the entire drawing is covered with tiny pixel holes, which are not noticeable at first glance, but if you start working with such an image, this can cause huge problems in the future. Again, everything is fixable, but this will require additional painstaking efforts to polish, so just like that "POP" and you will not be able to use the generated images.
And somewhy remix in Sora doesn't work for me. I paint the area I need to remake a write what I want and it still regenerates the whole image. But it works well in GPT chat.
Was your nose ring generated by AI? =]
11:03 LOL!! love how you gloss over the absolute slop you made as Mr. Beast quality. Samson you're a great channel but a trash designer
Is there any way I can use AI to Ask other AIs to make money for me? I want everything automated. AI art takes too much time to make a 1 minute video on youtube😕
I love gpt images but hate the victorian prude censorship
AI comfyui channel here 🎉
Revolutionary, Mind Blowing remarkably Insane. Completely game changer.
I like chatJtp and i am working on a music vdo using jtp image but its really censored. Can't do woman legs barefoot close up shot, cant do woman on a Beach in bathing suit that is to reviling anything that show too much skin or position that are not neutral like the ai told me cant do a couple hugging cant really swap face and I can go on. So im using chat but have to use other to get the things that chat as flag and don't let me generate. Don't know Why no one is talking about that.
Awesome ideas Samson, thank you. 🙂 Your black background makes me think, how much more black could this be… and the answer is none. =D
"Faux-tography"?!
Think we need a hashtag for that 😉
"…nothing like a massive beak…." almost spit out my coffee. Great video and tons of great ideas!
every new business model will only have a 30 second shelf life until it is replicated for free
Don't fall for it – openAI keep pulling the rug, so what works today doesn't work tomorrow. They have already done away with editing and gone back to 'recreating' the entire image. It's not trash, but it's nothing like what they launched.
Videos like these it freaking annoying… Claiming that it can just place your product/design on a mockup it a purely lie, AI cannot yet take images and place them on a mockup, it generates the image from scratch and therefore the design it not like it was original. Claiming this is untrue. I guess it will get there, but it's not there yet. For a this to work the AI needs to be able to generate an image and use an image you upload as is to incoporate in the generated image, and that is NOT how it works as of now
It's good to also mention weaknesses, in the first example of product placement, most of the time, the product has subtle differences in the AI generated image, even when giving it multiplr angles, these kind of small alterations to a product can be fixed in post, but sometimes it takes time. And when we talk about products the ai generated version needs to be on spot.
It was very poorly restored in Borja, Spain. Not far from where I live. 😂
18:39 to 19:05. Hehe.