ai image generators are not only evolving rapidly sometimes making images indistinguishable from photos but there are also new services that deserve your attention This may confuse you because the choice is getting too big Today I will show you five different realistic image generators You can try each of them for free In the first half of the video I will show you how to access and use each generator and then we will compare them side by side under similar conditions The first image generator I'm going to show you is called Rev Go to preview.rev.art to access it Links to all the services mentioned here will be in the description After creating an account you'll get into explore section If you open account settings you'll see that there are 100 initial free credits and you'll have 20 daily credits Each image costs one credit Generation process is very simple and really fast There is no different additional settings and tools You should describe what you want to see on the generated pictures as detailed as possible There are no predefined styles so you have to specify it in your prompt Here is aspect ratio adjustment You can generate up to eight images at once Prompt enhance will make your prompt more detailed which can make the result better especially if your prompt is short and too basic And seed value should work in the same way as in other generators like stable diffusion for example If I set some number images in different generations will be more similar Auto means that this number is different for each generation So probably there will be more difference between generated images Here are my images Double click to zoom each image and click one more time to see it full screen Okay probably this is not the best image I've ever gotten here but I like the lighting and overall atmosphere And there are three other pictures When some image is selected you can see two options Instruct and edit prompt If I select edit prompt I'll see the actual prompt this images was created with It's different from the prompt I entered because prompt enhance option is enabled You can change this prompt or some of these settings and create more images Instruct option allows you to specify which aspect of the image you want to change without the need to change whole prompt For example my instruction is make the image brighter but you should not expect that the consistency of the objects will be saved The difference between original and generated images is obvious Image to image feature works not like in other image generators You can just add your image here and then selected number of images will be automatically generated aspect ratio of the created images will be copied from the reference And if you previously entered some prompt and then uploaded image the prompt won't affect the generated pictures Okay here is another example with different reference image As you can see the dress on the generated images is not 100% the same as on the original picture And don't forget that if you want to save images in full quality you should zoom image and then use this download button The next service I want to show you is Runway which is known as one of the best AI video generators But there's also image generation tool On the free account you'll have 125 free initial credits so you'll be able to create around 15 images This generator has little bit more options than Rev Here I'll enter my prompt There are two additional features I'll show you how the references work in the next generation Now I'll select some of these predefined styles Aspect ratio adjustment Aesthetic range increases the difference between the generated images Here's seed value and you can generate one or four images at once Generation process will take longer than in Reeve AI Here is the result By clicking on image you'll see all the generation info and you can create variations generate video from this picture or add it to the references tab Here is how the references feature work You can add up to three images for example characters and places You can give some name to your character to save it and use in different generations Pictures with people can be used to create consistent characters and images with some places are for creating environment around the character from reference Since my character has a name I can mention her in the prompt This woman should stand amidst the highway during the storm I'll disable style preset Okay And aspect ratio will be 16x9 The result is pretty good All the generated images are really similar to the references The character looks like it almost the same person each time You can also use single image of a character and describe all the other details of the image in your prompt Okay for some reason this generation was blocked I set number of images to one and it helped Wow Honestly I'm impressed by this result I didn't expect that the character will look 100% the same as on the reference Google has two separate tools for image generation One of them is regular image generator image FX based on 3 model It isn't available worldwide yet so I use AmericanVPN to access it For example I want to get images of a pilot inside his Formula 1 racing car Here are additional words you can add to the prompt to create some styles and effects In the settings tab you can see the seed value which works a little bit different from other image generators and of course aspect ratio adjustment You'll get four images in one generation by default My images are ready I specified that this is Red Bull Formula 1 car And on the first and fourth images the Red Bull logo is clearly visible Yes the rest of the text on the car in almost all cases looks messed up but overall quality of the images is pretty good Nice Now I want to show you how to create consistent characters here First of all you have to provide description of the appearance of your character The prompt should be as detailed as possible I want to get images of a medieval prince All the generated images have something in common but this is not the same person So after creating description of your character you have to lock the seed value So the person on each image will be almost the same As you can see the result is more consistent Yes style of the mustache and beard is still a little bit different but in such cases detailed prompt can help or you can simply make few generations in a row and select the best results Second Google's image generation tool is whisk It has few additional features similar to what we saw in runway You can add image with some object or character picture with environment and style reference I'll leave style reference placeholder empty and I won't add any prompt because I just want this girl to appear on the image with storm and highway In the settings you can see the seed value and select model In image FX you can use only imagining 3 model but here's Imagin 4 available too which can probably create even better images than version 3 I almost forgot to mention that Google's image generation tools are pretty strictly censored This become obvious when you try to create images with women The generation can be cancelled for no reason Like in this case I received only one image instead of two I think in runway references option works better because it allows you to create consistent characters from the reference image But here the girl on the generated picture is not similar to the reference You can use refine option to change some parts of the image I wanted to make this girl smile Okay the girl changed pose but the consistency of the character on these two images is obvious Nice And what if I add one of the images I created in runway to the style reference section there's no so much difference but you can see the water drops on the body of the girl this time And if I'll try to use this knight as reference you'll see that night on the generated images isn't the same So this reference tool can be useful if the consistency of the objects is not really important The four realistic image generation tool I want to show you is Edog AI Here you'll have 10 weekly credits which means around 15 highquality images Here is creation tab and in the explore tab you can see images created by other users so you can reuse their prompts and other settings for your generations I'll enter my prompt All the generations are public on free account Here's aspect ratio setting The latest version of the image generation model is three When this model is selected the generation costs four credits If I'll select one of the older models it will be cheaper of course And here you can see your render speed setting which should be called quality adjustment because default means medium quality Turbo means lower quality but faster generation and the quality option will provide the highest quality but the generation will cost six credits in this case Magic prompt will enhance your image description like in other generators and obviously style will be realistic Plus there are some predefined color palettes You can create your own only with paid subscription Negative prompt and seed value editing options are unavailable on the free account too Okay let's finally create my images As you can see face and hands look really nice on the first image In other cases faces have some imperfections which is typical for AI generated images with human faces placed on some distance from the viewer So I'll try to make the character closer to the camera This time faces should look better First picture is just perfect as for me On the next two images the woman is obviously over 25 years old but overall quality is great Unfortunately imageto image option is also paid So that's all I can say about Ideogram AI And of course the most famous one among these services is Chad GPT40 Now you can create up to three images per day without paid subscription You can describe the image you want to get in the way you want or you can use some reference image with corresponding instructions The quality of the images created by chat GPT is really high But there's one thing I hate about this generator It's easy to spot which pictures created by chat GPT because there's some yellow tint Even after I asked it to make winter version of this image you can see that image is still little bit yellow Let me know in the comments if you have some ideas why this happens My theory is that this was made intentionally Though if you'll try to create fake image with real people or places and use it as is it will be obvious that this image generated in chat GPT But there are a few ways how to get rid of this yellow tint First of all you can edit already generated images but it can be hard to do if you have lots of different images You can get images without unnecessary yellow colors inside chat GPT using special words in your prompts I'll use the same prompt I used before but here I'll specify that color temperature is 9,000K The image is too dark But even now you can see that there is no yellow tint at all Nice If I'll recreate this image but 9,000K will be replaced with 7,000K Image is brighter And here's still no annoying yellow color You can search for color temperature charts Values higher than 6,000K mean cool light This method works good if you specify both the strength of the lighting for example bright sunlight and color temperature What about realistic character consistency in chat GPT i'll use the same image I used in Runway as character reference and I'll ask it to recreate this image but girl should look surprised Yes lots of details look the same but this girl is not really similar to the reference as for me Runway did better job Okay I'll make second attempt The girl should be in swimming pool and her face should be serious The face of the girl is just little bit similar to the reference Plus she looks like she didn't sleep all the night So probably chat GPT will work better if you'll create some cartoonish consistent characters with lower amount of small facial details as I promised Now I'll compare all five mentioned generators in similar conditions First let's see how these services will create extreme close-up portraits I selected the best image from each generation Google's image generator did the worst job It's pretty easy to spot that this is AI generated picture I created it with Imagen 3 model because Imagin 4 generated even less realistic images Chat GPT4 a created better image Skin texture and face details aren't too sharp and flat like on the previous image I cannot find any obvious artifacts or imperfections on the image created in ideoggram AI All the details and lighting look cinematic Image created in runway also looks great Character looks like this is real person His eyes don't look into nowhere like on the picture created by chat GPT Two previous results were great but probably Riv created the most detailed and realistic skin texture The only thing is that this defocus effect on the nose could be more realistic as for me So I definitely can say that ideogram AI runway and Riv are the best for creating realistic close-up portraits But lots of generators have the same problem with creating faces on some distance from the camera Here's the result Ideogram AI created image with obvious creepy artifacts on the woman's face Fingers don't look good too Image by chat GPT doesn't look really bad but when you zoom in face looks like it's some drawing face on the image created with runway is a little bit unclear and here are problems with nails on both hands riv created almost the same imperfections on the face and hands but image doesn't look completely spoiled Google's image in three not bad I should say at least I'm not scared looking at the woman's face as you can see all the generators has the same problems with creating clear faces and hands when the person is not really close to the camera but probably chat GPT He did it better the other four services Now let's see how each service will generate hands I asked each one to create images with five hands forming a star shape Only image created with Google's imagin 3 has missing finger failins In all other cases hands look nice Chat GPT and Riv actually created five hands forming a star shape But you should understand that none of these generators can create hands with right number of fingers all the time For example this image by chat GPT is my third attempt to get what I want Another prompt I want to test is front view of a man in the forest waving with both his hands I'm really impressed to be honest All the services did pretty good job I can't say anything bad about any of these images The fourth test is to compare how each generator will work with complex prompts when there are lots of different objects on the scene Here's my prompt Images should have two guys and two girls sitting on the beach around bonfire Girls should wear sundresses One guy holding a stick with marshmallow The sky is a little bit pink and orange and it's getting darker and some stars are visible now and there should be ship on the distance in the water Runway created almost perfect image All the elements I mentioned in the prompt are here There is just additional stick with marshmallow in the bonfire Chat GPT also did good job There's no unnecessary or missing details Ideogram created additional sticks with marshmallow and second ship and all the faces are spoiled at all because people aren't close to the camera Riv created not bad image in terms of prompt correspondence I like that there's some action in this scene but faces aren't really clear And I used Image and 4 model to get this image For some reason all the people holding sticks with Marshmallow but it's not bad result at all I think the best images created chat GPT and runway These images have the best quality and they correspond my description In the next test I'll try to create images with real people real existing brands and objects like cars of some brand and of course existing places For example I tried to create images of Tom Cruz wearing black suit standing amidst the desert Pictures created in Riv and Idog look really great Runway and Google's image generation tool didn't create any images because such prompts violate their content policies The next example is Ford Mustang branded with Coca-Cola and Microsoft logos I didn't specify if it should be new versions or retro car So it's different on different images The only generator which has partially failed this task is runway Here's Mustang with Coca-Cola logo and colors but I didn't get any image with Microsoft logo in this generator All the other services did pretty good job this time All the things I mentioned in my prompter here And what about real places new York or Paris probably will be too easy So I decided to create images of Ukrainian Lviv city Here's how this city looks in reality architecture is pretty similar to what you can see in Poland or Czech Republic Ideogram created the right atmosphere and style of the architecture but you won't find any of these buildings in Lviv Chat GPT on the opposite tried to create the real buildings you can find in this city but overall quality is not really high Image by Google's Imagin 4 model is also really close to what you can see on the photos Quality and small details are much better than on previous picture by chat GPT I got not bad result in runway It's not 100% similar to actual place I wanted to create And if you zoom in you'll see that the windows and other details of the buildings are unclear And probably I can say the same thing about the image I got in Riv It's a little bit hard to choose the winner because Chad GPT recreated real existing buildings but the quality is lowest among all five results Google's image generator is on the second place in terms of accuracy but the quality is much higher than in four other cases All other services created right architecture style only Let's summarize everything we've recently learned Google's image generator has probably the most generous free plan because it doesn't have any strict limits on number of images you can create per day The only services which allow you to create consistent characters are runway and image generation tools from Google If you ask ChatGpt to create image with the same character as on the uploaded picture the result won't be really consistent Google's image generator allows you to create consistent characters only with prompts So you can't use some image as characters face reference As you can see almost all the services still have problems with creating human faces on some distance from camera Surprisingly the only generator which partially failed task with hands is Google's image generator I tried to use third and fourth versions of Imagin But the best image I got still had problems with fingers All the services did pretty good job when I tried to create complex scene except ideoggram AI which multiplied different mentioned objects and faces of the people were completely spoiled No one failed the test with real places and brands but there are restrictions on creating images of real people in Google's generator and runway Chad GPT could become winner in this test but the quality of the generated picture of Leviv city was much worse than all the other results In conclusion I can say that it's hard to find perfect universal service for all possible cases But I hope this video helped you to understand the difference between these tools and define which ones are suitable for your goals