a new day a new video and we have something new to learn in this video we would go over on How to Train chat GPT 4 or 3.5 whatever you have to create prompts for your AI diffusion platforms like stable diffusion mid-journey or any other platform that you use we would focus only on how to create street photography portraits with AI without using an expensive DSLR or the high-end mirrorless cameras you find nowadays mind you these are not the same as Studio portraits we would cover that in another video by the time we finish the video you will learn one some basic concepts of photography two some concepts related to Street portrait photography only three using this knowledge to generate your AI portraits don’t worry your AI skills we’ll always try to simplify things as much as possible and focus on applying these basics of portrait photography into Air Generation how you use these skills is totally up to you maybe this would help you improve taking better shots from your iPhone or your triple camera android setup or you can just use mid-journey so enough Chit Chat let’s dig into the video [Music] yup don’t ask the Firebox is only for this video I stick to chrome we have our chat GPT saved in the bookmark and oh well history is not available even for the plus version at the moment let’s search wait what what the do you fashion model what’s the fashion model let me do a quick search on what do fashion model is Wikipedia Wikipedia Wikipedia Wikipedia Wikipedia Wikipedia Wikipedia Wikipedia Wikipedia Wikipedia there you are what the image in that space of all images what is all this why am I wasting my time when we have mid-journey this is how we train chat GPT to generate prompts Ty do you know what a diffusion model is don’t explain please answer yes or no until stated otherwise do you understand [Music] laughs then type the following [Music] [Music] now I need to explain to you some things the text within the brackets is what the AI generates randomly or as per what parameters I have defined however the texts outside the brackets is fixed in the prompt let’s understand slowly and carefully portrait of the is what is fixed the AI may use a male or a female randomly in the prompt then I asked the AI to Define five random adjectives which can be anything for example engaging humorous sociable ethereal enigmatic dreamy etc etc you get the point let us understand some camera Basics here before we move on to why we use these specific settings for a portrait or candid street photography the camera model the AI here will randomly choose a camera model from Nikon Fuji Canon Leica Etc we use the word close up as we want a full headshot or at least a shot covering from the chest till the head the lens focal length is the actual distance in millimeters between the optical center of the lens and the camera’s sensor in simple words what I mean is the higher the focal length the more zoom you have and the lower the focal length The Wider shot you get that’s all you have to remember in street photography 35 millimeters or a 55 millimeters focal length is The Sweet Spot photographers mostly use 90 millimeters lenses for weddings when they need to take a shot from a distance in every diffusion model you can Define the type of lighting you want for the environment or subject for this formula we are only asking the AI to randomize between three styles of lighting as I publish more videos on YouTube I will Explore More styles of lighting suiting the subject of the AI generative artwork aperture is how wide the diaphragm of the lens is open for light to pass in in simple terms just remember the lower the number the higher the aperture meaning more light can pass in for example when you are shooting street photography at night with a lens having an aperture also called f-stop of 1.2 you can take photos handheld and with minimal noise compared to photos taken at lower apertures typical apertures for a lens can go as high as f-stop 0.9 to as low as f-stop 22. the higher the aperture the more expensive the lens and vice versa for night street photography choose an s-top of 1.2 to 1.8 if you are shooting handheld for daytime you can use anything depending on shutter speed the composure of the shot and the type of photography I will not go into much details regarding shutter speeds as it is not relevant to this video maybe in some other video I can try to experiment with it also the higher the f-stop better the bokeh bokeh is the visual quality of the out of focus areas of a photographic image I have put one location around the world which actually makes the AI randomize a single location for the shot you can however Define it as in the USA or in France for example as well more on that later in the video yeah I know I use 32k twice that’s a typo so ignore you can put 4K 8K or 32k they normally use these in prompts to define the details of the output as they are resolution sizes [Music] sharp focus is self-explanatory I prefer to use it in facial close-ups there are different type of engine renders that diffusion models use for this video we would stick to Unreal Engine 5 or octane render only photorealistic is a keyword I normally use when I want realistic artwork no matter the subject now we tell chat GPT to create seven props for us [Music] [Music] while creating this video chatgpt was extremely slow in response it time now plenty of times and I had to retake the scene several times [Music] if I had a singing voice I would sing to the seventh prompt but I’d better keep quiet my wife complains a lot when I am bathing [Music] I am not fast forwarding the video so you can see how much time chat GPT version 4 takes these prompts are a bit detailed so be patient I will write all the prompts in the description below prompt 5 prompt 5 prompt five prompt five prompt five prompt five prompt five prompt five prompt five prompt five prompt five prompt five prompt five prompt five prompt five prompt five yes [Music] [Music] after the seventh prompt finishes I will showcase plenty of images that got generated based on these prompts after that I would tweak the prompts to show you how you can create some variety in the Air Generation and get what you desire these prompts are used with mid-journey version 5. [Music] finally the last prompt [Music] look at these photos which were generated using mid Journey version 5. they are perfect one of the photos had some text marking at the bottom which I will show at the end of the video on how to remove them using Photoshop the prompt 2 photos are very realistic they are accurate as for the specs defined in the prompt with regard to focal length and aperture the price of a Nikon d850 ranges from 1900 to 2500 USD [Music] in the third prompt the AI generated a photo based on the Sony a7r4 camera which costs about 3 200 US dollars however there is no lens at a focal length of 90 millimeters with an aperture of 2.2 Sony officially only has a 90 millimeters 2.8 macro lens this is the magic of AI art generation where you can mix and match specs that do not actually exist in the market [Music] The Prompt 4 images are beautiful you can tweak the prompt to change the subject’s hairstyle hair color ethnicity Etc remember chat GPT only assists you with the prompt and you can be as creative as you want the level of customization is literally Unlimited [Music] for prompt 5 the AI generated a black and white image in my opinion it looks good you can however tweak it to get a colorful picture just add keywords like vibrant and colorful [Music] look at the details of both images generated by prop 6. if you find the bokeh too strong or dreamy you can always play around and reduce the depth of field [Music] look [Music] I had to make a variation of the second image in prompt 7. the details were not just up to my mark the variations worked remember that when you see an image check it for mistakes and try variations sometimes the variations will give better results and sometimes you must regenerate the whole prompt or write a new one thank you mid-journey or any other diffusion model is not perfect you must pay attention to details and may not get what you want on the first try let’s talk about the mistakes look at these photos I don’t know what I am doing wrong or if it’s just the AI it was impossible for me to create a rain image with people holding umbrellas [Music] after plenty of attempts I defined the keyword holding an umbrella in hand still I could get only one image correct if you guys have figured out how to get a perfect realistic photographic image of a couple or person holding an umbrella in the rain let me know in the comments now let’s talk about the hands one of the major improvements in mid-journey version 5 is that it got hands correct however there are still glitches and it’s far from perfect look at these images and look closely at the hands they are missing distorted or just not right I could not get the hands correctly in many images especially couples holding hands I tried variations for this particular image as you can see I can’t get their hands right I just scraped The Prompt and created a new one hands are far from perfect [Music] you can use these prompts tweak it and train chat GPT to generate prompts for photographic realism in a futuristic imaginary world check out some samples [Music] [Music] some of the photos generated by mid-journey version May 5th have some weird text either the bottom right or the bottom left [Music] in this case you can open the photo in Adobe Photoshop then go and select the healing brush tool you can see how it works so now with your brush let’s zoom in now just swipe it across and voila the photo is fixed [Music] to conclude AI is as good as your imagination whether you use chat GPT or not your creativity and knowledge will create beautiful Street photographic images even in chat GPT you have to train it a massive difference exists between using a formula to train chat GPT and simply asking chat GPT to generate a prompt you can use many AI platforms but mid-journey is one of the best for the art generation maybe in the near future we will get something better than mid-journey companies like Adobe have also announced their own AI generation platforms so let’s hope for the best your AI skills only gives you in-depth knowledge about a particular topic using it is up to you work play or fun our main goal is to teach everyone to use AI because one way or the other you have to hop on board the AI train I am leading the seven prompts text in the description so you can play around with whatever diffusion platform you use and also I am leaving a huge zip Google Drive Link which you can download all the images showcased in this video I hope you have learned something by watching our video we are trying to keep it as simple as possible let us know what you want us to focus on in the next video we love to hear your comments and input until next time if you have watched till here thanks please like And subscribe and hit the Bell notification to receive a notification as soon as we upload a new video thank you