“In this video, we’ll look at how well the AI understands photographic language and how realistic you can make the AI photos. I’ll give you tips on how to make AI photographs look realistic and what prompts to use – and most importantly what prompts not to use.
I compared Midjourney testp, Midjourey version 4, Dall-E 2, and Standard Diffusion to answer the Question: Which AI image generator makes the best photos?
This video is all about text-to-image prompts. I have a Separate video on how to do image-to-image (img2img) generations and how to use photos as part of your prompt.
[Music] Hi I’m Jen with making the photo as a Photographer I make my photos using a Camera and a photo editing program but Is AI about to replace your camera Should you sell your camera equipment In this video I create photos in Midjourney Dolly and stable diffusion I Look at how well the AI understands Photographic language and I show you how Realistic you can make an AI photo along The way you’ll find out which AI makes The best photos when I made this video Mid-journey was Alpha Testing version 4 But they also had the test P or test Photo model available I looked at both For comparison I also ran prompts through stable Diffusion and dial E2 to see which AI Makes the best photographs this video is All about text to image generation in a Separate video I’ll cover image to image Generations and how to use actual photos Or images as part of your prompt let’s Start with the obvious if you want AI to Make you a photo make sure to ask it for A photo use words like Photo photography Or photograph as part of your prompt Just saying realistic or real world Isn’t enough then add photographic terms To your prompt using photographic Language makes it more likely that You’ll get a photo realistic image I’ve Published a lot of Articles defining
Photographic terms and how to create Photographic effects both in camera and With photo editing programs if you want More information on photographic terms I’ll put a few links in the description Below let’s go through some types of Photographic terms you can use even if You aren’t making a photo some of these Turn terms may be just what you need to Get the look that you want first don’t Just ask for a photo as for a specific Type of photo you can ask for food Photography a still life photo or street Photography prompt for a wildlife photo A sports photo or photojournalism you Can ask for portraits glamor or fashion Photography landscape photography is Popular but you can also ask for fine Art or conceptual photography these Styles are more artsy than real some of These genres are more recognizable to The AI than others if you have a Photography background prompt for Specific cameras and other photographic Equipment you can ask for drone photos Or images from a specific type of camera Like a Canon Nikon or a Leica you can Ask for infrared photography Or you can ask for Polaroid shots or Mobile phone photography mid-journey Version 4 even knows how to make a photo Shot through a reflecting ball it’s not Exactly right but it’s a pretty cool Effect you can prompt for specific types
Of lenses like a fisheye Or a macro if you’re getting too much Blur in a photo especially a macro photo Ask the AI to make the image in sharp Focus or make it Focus stacked you can Even try asking for specific lenses from Canon or Zeiss or specialty lenses like A lens baby the AI may or may not know The specific type of lens but asking for Specifics emphasizes that you want a Photograph the AI is particularly good At creating historic photography and Film types like wet plate tin type or Cyanotype you’ll get the Styles colors Tones and artifacts found in old photos And some AI can render specific film Types like Kodak ectochrome or Fujifilm Velvia Also prompt for specific camera effects And settings by camera effects I mean Visual effects that are created by doing Something with the camera like Intentional camera movement Or panning or creating a zoom blur You can ask for a long exposure effect To blur the water in these waterfalls or You can ask for High Speed photography With a fast shutter You can either ask for Sharp Focus or Bouquet with a shallow depth of field You can ask for lighting effects like a Lens flare or light trails You can even ask for more Advanced Techniques like double exposure or HDR
Photographs you can definitely prompt For f-stop focal length and shutter Speed this emphasizes the photography But I’m not so sure that I’m seeing the Expected visual effects in the image Photographic composition isn’t Necessarily different from other Visual Arts but there are some compositional Techniques used a lot in photography You can ask for negative space or Leading lines You can ask for a low vantage point or Aerial views You can even prompt for forced Perspective shots or specify your Foreground mid-ground and background you Can use compositional terms like rule of Thirds symmetry or golden ratio but I’m Not always sure the AI gets it and Mid-journey got a little confused when I Asked for framing a subject framed by Another element He gave me both the moon Framed in the tree branches and the Entire photo in a frame if you’re after A portrait or something like still life Ask for specific lighting you can ask For natural light or specify the light’s Position in this case behind the person Creating a silhouette or specify a Specific type of Studio lighting like Split Lighting in this case stable Diffusion did get a little confused and Split my image too you can ask for Cinematic lighting Rim lighting or a
Catch light in the eyes you can even Prompt for special photographic editing Effects like you can do in Photoshop for Instance you can ask for a selective Color or a vignette of course you can Create these effects later on any image And I’ll put a link to a couple of Tutorials in the description below or You can ask for a tiny planet photo this Effect is pretty easy to achieve in Photoshop with any photo as I did with The selective color image I’ll put a Link in the description to a quick Tutorial on how to make tiny planet Photos in Photoshop like other forms of Visual art asking for a photo in the Style of a famous photographer is a good Shorthand way of telling the AI how you Want the photo to look I’ll include a Link to the 20 most famous street Photographers you should know but look Around for photographs you like there’s No guarantee the AI will recognize the Photographic style but the images may Give you some good ideas mid Journey Version 4 nailed the Styles and the Subjects of Vivian Mayer and Elliot Irwit who is famous for photographing Dogs of the streets of New York City and Mayor who took a lot of selfies in Reflective surfaces with her brownie Camera these AI images seemed so real That I actually ran a reverse image Search to make sure it wasn’t just
Pulling photos off the web any Leibovitz And Lindsay Adler are well-known Portrait and fashion photographers with A distinct style but you can prompt for A more generic style put in a Publication like Vogue life or National Geographic which all have their unique Photographic Styles including the words Award-winning photo in the prompt May Improve the image quality or ask for a High resolution photo if you’re trying To create a realistic looking photo There are a couple of terms you should Avoid the term photorealism seems an Obvious choice if you’re trying for a Real photo but photorealism is a style Of painting and sculpture the AI May Ironically return a less real photo if You use this term though not always also Generally avoid terms that refer to Video game Graphics or other graphic art Forms terms like octane render Unreal Engine 8K 4K that sort of thing But as with all rules there are Exceptions The AI does sometimes return photos Within photos if this happens try Changing the word photograph to photo or Asking for no border I ran a lot of Photography terms through the top AI Image generators and I have a few Thoughts on which is best for creating Photos but first let’s see what you Thought a couple of weeks ago I put a
Poll up on my Community page with four Cartier person style photographs all the AI generators did a good job recreating This iconic photographer and the Decisive moment the results are in you Thought doll e was best after running Nearly 100 photographic terms and styles For the various models I ended up liking Mid-journey version 4 the best Mid-journey test photo came in second It’s actually easier to get photographs In stable diffusion and Dolly but the Photos aren’t necessarily that great They are more like quick snaps taken With a mobile phone mid-journey makes More artistic Fine Art photos Mid-journey versions 4 has much more Photographic knowledge but it will try To go painterly on you when it can Especially on landscapes and portraits Probably because there are so many Paintings of Landscapes and portraits Out there but you will always get a Beautiful image out of mid-journey Whether it’s photographic or not while People aren’t perfect in mid-journey Right now they are less distorted than They come out in stable diffusion and Dali can you make a photograph without a Camera the answer is yes you can make a Photograph with AI use Photo terms Describing lighting focal length or Camera settings AI won’t replace the Experience of a photo shoot but it can
Create stunning photographs let us know Your best photography prompts and try Out any photographic terms and styles That I missed if this video was helpful Remember to like the video and subscribe To the channel This is Janet making the photo let’s Make something amazing together [Music]