DALLE-3 Masterclass: Everything You Didn’t Know (Complete DALLE 3 Tutorial)
Vložit
- čas přidán 14. 06. 2024
- DALLE 3 is incredible and in today’s comprehensive DALL-E 3 tutorial, we’re going to cover ai image prompting, DALL-E Vision, image re-imagination and analysis, using GPTs with Dalle-3 and so much more.
DALL-E 3 is powered by GPT-4. Which is the incredibly powerful large language model that also powers the latest version of ChatGPT.
And OpenAi’s research showed that using very detailed prompts gives significantly better results when generating images which is why DALLE 3 has the amazing ability to rewrite prompts to generate stunning visuals.
Prepare to level up with this full DALLE-3 course.
#dalle3 #chatgpt4 #aitutorial
Chapters
0:00 Getting Started
1:35 DALLE 3 Prompt Basics
4:18 ChatGPT to Prompt DALLE
6:11 Editing Images
8:41 DALLE 3 Aspect Ratios
11:16 Text Generation
13:09 GPT 4 and DALLE Vision
15:19 Image Analysis
16:03 Image Re-imagination
17:25 Build DALLE GPT
23:25 Custom Instructions
24:43 Limitations
26:05 Key Takeaways - Jak na to + styl
SUPER AWESOME TUTORIAL! Especially the section related to creating custom GPTs! I will definitely be sharing your channel with friends! Looking forward to more!🎉
Thank you so much for the positive feedback! Really means a lot as a new creator!
Such a detailed video. Thank you for this.
Hey! Great video. Subscribed and looking forward to more. Thank you.
Thanks for explaining and teaching in detail about the Dalle and gpt
This was good and well explained, thank you
Thanks! Just the tips I needed.
Perfect! I’ve already created a few GPTs, but I’ve been stuck on adding more custom instructions. Thank you!
Glad to hear!
Thank you for helping us understand the amazing AI tech that is out there for us all. If it was not for guys like you creating these videos, the rest of us would surely miss out!
Really appreciate the awesome feedback!
Really love this video ! love how you explain things, it gives me the vibe of being in a classroom you explain everything at the right pace.
Subscribed and def going to be watching you more!.
Thank you again fam
Thank you so much for the awesome feedback! Really means a lot to a new creator.
you're very welcome keep up the amazing work mate.
@@AIcents
Great video! You've convinced me to try out Dall-E for artwork/thumbnails for my podcast. Thanks for your help!
Glad to hear and good luck with the podcast!
Awesome Tutorial! Thank you very much! You earned my sub
Thanks for the sub and feedback, much appreciated!
Great tutorial - thank you.
Glad it was helpful!
Fantastic video !!!
Thanks, really appreciate the feedback!
I want just say “ thank you so much ❤ “ for this video, keep up
Glad you found it valuable!
💪Great video
I use gpt4 a lot when mastering dungeons & dragons. I make NPCs on the fly, and when I get a version I like, I aks for the seed number and then proceed to get that same character in different situations / poses.
Also one thing DALL-E3 can do which Midjourney can't (well I tried I didn't get it to work) is mixing two characters in a picture like "a huge male viking brawler embracing a bald yuan-ti warlock with tattoes all over her body".
Great tips, thanks for sharing!
Cost?
thanks
This was so helpful. I have created a couple of bots but now I see how I can make them more tailored to my needs.
Really glad to hear!
🎯 Key Takeaways for quick navigation:
00:00 🚀 *Introduction to DALLE-3 and Getting Started*
- Overview of DALLE-3 tutorial covering prompting, DALL-E Vision, and image re-imagination.
- DALLE-3 is powered by GPT-4, requires ChatGPT Plus or Enterprise subscription for full features.
- Accessing DALLE-3 via ChatGPT window or Explore page, ensuring plugin settings for usage.
01:22 🖼️ *Image Generation with DALLE-3*
- Detailed prompting enhances image generation results with DALLE-3.
- Utilizing GPT-4's prompt rewriting for optimized prompts.
- ChatGPT as a valuable brainstorming tool for generating image prompts.
06:00 🎨 *Editing and Refining AI-Generated Images*
- Refinement process for AI-generated images through iterative prompting.
- Modifying prompts for desired image outcomes.
- Utilizing aspect ratio settings for different image formats.
11:08 📝 *Text Generation with DALLE-3*
- DALLE-3's ability to generate legible text within images.
- Addressing typos and optimizing prompts for text generation.
- Considerations for incorporating text in DALLE-3 generated images.
13:30 🍳 *Utilizing DALLE-3's Vision Capabilities*
- Leveraging DALLE-3's computer vision for image description and analysis.
- Practical use cases including recipe suggestion, artwork analysis, and image re-imagination.
- Exploring the capabilities of DALLE-3 for diverse visual tasks.
17:44 🤖 *Building Custom GPTs with DALLE-3*
- Introduction to building custom GPTs integrated with DALLE-3 for specific tasks.
- Configuration steps and capabilities selection for custom GPTs.
- Demonstration of creating a custom GPT for aiding in image generation workflows.
19:39 🛠️ *Building Custom GPTs*
- Creating custom GPTs allows for highly tailored AI assistance,
- Instructions can be specified to guide GPTs in generating responses according to user preferences,
- Custom GPTs streamline workflow and can be easily saved and shared for future use.
21:35 🎨 *Brainstorming with Custom GPTs*
- Custom GPTs can assist in brainstorming by generating ideas based on user-provided prompts,
- Users have the flexibility to guide the level of detail in prompts or let the GPT generate ideas independently,
- Generated ideas can be refined further based on user preferences before finalization.
24:55 🚧 *Understanding DALL-E 3 Limitations*
- DALL-E 3 has limitations such as maximum prompt length and copyright guardrails,
- Replicating living artists' work is prohibited due to copyright laws,
- Users need to be cautious of evolving limitations and potential challenges, such as rendering realistic human hands.
26:23 📝 *Key Takeaways for DALL-E 3 Journey*
- Specific and detailed prompts yield better results,
- Iterative approach is recommended for refining generated images,
- Setting aspect ratio in the opening prompt can optimize image generation process.
Made with HARPA AI
Awesome!
Thanks, glad you enjoyed!
Fascinating, I hadn't realised Dalle3 was so limited. The images seem very oversaturated and derivative. More of a toy than tool as you seem to have little control over the fine details.
+1
Good job. I've got it right myself, just wanted to see what's new. Hiccups are in place, still.
Fantastic timing - ive been using DALLE3 on and off but many of your tips of wisdom I was unaware of so this masterclass was perfect to get more up to speed!
One thing I cant seem to figure out is how to get photoreal images as most of the time regardless of what instructions I give in prompts, including those provided by DALLE3 such as "ultra-realistic, high-definition photograph" the images whilst beautiful more resemble 3d Pixar or 3D Game style rather than real looking photos - are there any tips you and your audience might give on some useful keywords to get closer to photorealistic?
Glad you got value from the masterclass! As for photorealistic images, DALLE is a bit more finicky compared to other tools such as Midjourney. However, you are on the right track with keywords such as 'photograph / photo realistic / real life'. Ultimately though, it will still involve a bit of tweaking while being detailed as you can in your prompt in terms of composition, lighting, etc. to really hone in on a realistic looking generation.
Waiting list to subscribe is the killer!
It really is! Fingers crossed the wait won’t be too long.
Amazing Tutorial. How do i upload one of my own photos (profile of me) to change it to an AI image?
Great content thanks. Have you tried Runaway to extend a 9:16 to 16:9 instead of regenerating it? I get better results that way.
Thanks for the tip! Hadn't used it yet but will definitely give it a try.
SIr , is it work with eommerce business ??
Well scripted and produced video 👍
Thanks, really appreciate the feedback!
how do you get Dall-e to refine artwork that I like?
I find that once I really enjoy elements of a picture or theme it will not keep those elements if I want to refine other things like backgrounds. Your advice?
That’s the tricky part with the current version of Dall-e. Usually best course of action is too keep refining the generated prompt of the image variation you prefer the most. Can be a bit painstaking at times but can also be worth it depending on the situation.
I work with dall e alot, but their freaking policy is pain in the ass and their greatest down fall if you ask me. and the lack of AR is also such a back lash, do you have any tip about the portrait fail ? many times dall e makes my promps that I ask in portrait I even say 9:16 but it keeps making it portrait horizontal like landscape
Excellent video. I tried Dalle out yesterday, and I admit it's pretty amazing. However, I could never get Dalle to recreate characters I liked. Dalle could not take a character and change his clothes from casual to a business suit, nor could I tell Dalle to make the character face 20 degrees to the right. I ended up with a slightly but noticeably different character (a brother or cousin, if you will). It's of little use to me until Dalle can recreate a character in different poses and wearing different clothes. I think that Dalle needs a way to assign a drawing a GUID, for example, and be able to use the GUID to recreate a drawing with a specific modification. Perhaps paid users would get so many GUIDs that would always be available?
Character consistency does seem to be Dalle’s Achilles heal. Though the dedicated GUID would be a really compelling feature if it was ever offered down the line.
😍
I tell Chat GPT what I want and then have Chat GPT write the prompt for me. Then I’ll make small changes to the prompt, if necessary and then copy and paste the prompt for DALL-E to create, lol.
niceeeeeeeeeeee
Hi, Could you please tell me how much this costs. I would really appreciate it.Your video is fantastic. Well done.
Glad you enjoyed and really appreciate the feedback! The cost to access all these features is $20/month as part of the ChatGPT Plus subscription tier. But there currently may be a wait list to get access as OpenAI is struggling to keep up with demand.
Can I prompt "women who fight crime dressed in cheerleader outfits"?
Sorry my friend . You not use dalle3. Just try it. Ask your assistent which dalle version you use ? It will says dalle 2 . Becouse you didnt set it on screen as option “dalle 3 beta” becouse you cant see the screen. Same problem at me
Can Dall•e replicate actual people from supplied images?
In a way yes, but also not really. You can give it an image of a person and ask it to recreate it but it'll generalize the person's features. So the new image won't really look like the person but maybe a distant cousin of sorts. Mid Journey tends to do this sort of task much more accurately.
We cant edit image…
The image resolution is bad, blur everywhere.
Muddled.
Thanks for watching! Feel free to share specific areas of improvement in the video. Everyone’s feedback is valuable as it allows me to better create videos that help supercharge people’s productivity and creativity with AI.
Who else tried to blow one of the lines of the screen
I don"'t understand how you make realistic pictures with DallE. They all look fake compared to midjourney
chatgpt and dalle3 are JUNK. 40 image gens every 3 hours? for 20 bucks?!?!?! no.
Midjourney images are much more realistic than DALLE images. Its not even close.
Indeed. Though ease of use over midjourney is what draws most to Dalle. But that may change very soon with the web version Midjourney is testing.
I found that the generated images are always the same style and boring. Even when you ask it to look like Rembrandt or Velasquez it gives you all this purple/orange crappy unrealistic and decorative look. POS I would say. Good for kids never drawn anything in their life. Awful for advanced or skilled artists trying to brainstorm ideas
What would you suggest then?
@@aleeez007 Depends on what you’re expecting. In my opinion the best thing to do is to create your own gpt using this one as a template, understand how this works and teach it to do something different. In theory. Otherwise look around and use mid journey, it looks more refined and much more tuned to give realistic results. It’s up to you what do you want to do.
Could it be due to the prompts being the same style and boring?
@@matt818 nah. Same prompt for midjorney gave much better and more realistic results. It could be the algorithm behind the scenes.
User error