DALL E 3 initial impression, Cons vs Pros

แชร์
ฝัง
  • เผยแพร่เมื่อ 5 ก.ค. 2024
  • #dalle3 #stablediffusion #aiimages #a1111 #aiart #dalle #chatgpt #midjourney
    Update: Dalle 3 is currently free to use as part of Bing chat features.
    00:00:00 introduction, will Dalle 3 replace stable diffusion?
    short answer is no, each has its own use cases
    00:00:043 Dalle 3 image generation features
    00:05:24 Comparison with Stable diffusion images
    using same prompts to generate sample pictures and comparison between Dalle3 and stable diffusion output
    00:08:15 conclusion
    So will Dalle 3 replace stable diffusion?
    despite that initial review of the AI Image generator, shows its superiority in composition, prompt understanding, level of details, context comprehension, and its various features, in my opinion, the answer is NO! because it is not open source, not free, and have many limitations in creating art, training your own subjects, cannot represent public figures…etc, so stable diffusion is more likely to catch up with Dalle 3 in the future and still used by most of the community.
    Lets see a short review of Dalle 3 and my initial impression of this very important AI Image Generation update that will become available next month.
    One of the biggest problems in Midjounry or stable diffusion is that it does not adhere to the prompt accurately and it drops many words from the description.
    We don’t see many of the tokens we describe in the prompt to be part of our final picture
    Dalle 3:
    Understands context very well, and a lot better than other AI Image generations in comparison, it understands prompts better and adheres almost to each text phrase
    Accuracy text inside images
    Works in conjunction with Chat GPT, tweak the prompt with Chat GPT
    Limitations (adult content, violent, doesn't represent public figures, training…etc. )
    At the time being, it seems Dalle 3 is more powerful than all other AI Image generators, but it has limited use cases, it is good for commercial users who want to create safe for work content , stories, with high level of details, guided by power of Chat GPT
  • วิทยาศาสตร์และเทคโนโลยี

ความคิดเห็น • 19

  • @CHARIOTangler
    @CHARIOTangler 9 หลายเดือนก่อน +8

    As long as companies like openai censor and monitor the output of their generative ai programs, open source will always be superior.

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน +2

      agree.

  • @DanielSchweinert
    @DanielSchweinert 9 หลายเดือนก่อน +1

    SD should just assimilate Dall E3 and Midjourney so we can use it on our own and also completely offline :-)

  • @Mimeniia
    @Mimeniia 9 หลายเดือนก่อน +2

    Impressive but not for the mainstream user who expects unlimited, ungated features. If it can handle multiple characters properly then i'll be jealous.

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน +1

      exactly, it is amazing, but wont be used by most people with current cost and limitations.

  • @alandonenfeld2070
    @alandonenfeld2070 9 หลายเดือนก่อน

    How do I contact you for a project?

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน

      I am sorry, not taking private or business contacts at the time being.

    • @alandonenfeld2070
      @alandonenfeld2070 9 หลายเดือนก่อน

      can you recommend anyone?

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน

      sorry, I can not.

  • @thegreatdelusion
    @thegreatdelusion 9 หลายเดือนก่อน

    Hey, comparing this to Stable Diffusion isn't fair. You gotta prompt SD differently and add way more details about the image like lighting, if it's a photo or animation, etc. The models you used are also fine tuned towards characters specifically. You can get images that are just as good as DALL-E 3 if you just use better prompts.

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน +1

      SDXL can generate very impressive pictures and adheres to large part of the prompt as per the video samples, but not as the level of details in the samples provided by Open ai Dalle 3, SDXL better prompting can create stunning pictures... either ways, I still believe, that SDXL and Stable diffusion will remain more superior in terms of adoption and continue to improve, because they give us ultimate freedom in creation and we dont have to put our data on someone Elses servers and trust that they will not do anything with it.

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน +1

      the big difference between the two is how good Dalle 3 is following the prompts and composing the elements of the prompts inside one coherent picture.

  • @SEKOPASA
    @SEKOPASA 9 หลายเดือนก่อน

    Why would a paid censored service be end of a free, uncensored, open source, community driven application? Please try to be creative rather than click baity with your titles.

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน +1

      hmm... you might have a point, I will adjust the title...some people dont know that it is censored, paid...etc. in terms of content, but the title as you have said, might not have been a good choice... adjusting.

    • @SEKOPASA
      @SEKOPASA 9 หลายเดือนก่อน

      @@AI-HowTo with this answer you have gained my respect and a humble follower. Thank you for listening.

  • @alexlefkowitz
    @alexlefkowitz 9 หลายเดือนก่อน

    Hey how can we contact you for business inquiries?

    • @AI-HowTo
      @AI-HowTo  9 หลายเดือนก่อน

      I am sorry, I am not available at the time being for private or business inquiries.