By -sp●●n-, 1703891303
I wanted to show how AI can be used in a positive way, to quickly work an image. I had this image taken a while back but never did anything with it, the bar to entry to fix it was too high for me (10 years back), would likely have taken 2 days back then to get something decent, the image was a good pose, slightly blurred (straight off the camera, including light setup slightly visible):
First step was to correct tilt, perhaps I was drunk that day (simple image >> rotate >> free >> 5 deg clockwise), then crop in to just the model. Next up was to grab the model from the back drop, 10 years ago it would have taken much longer background eraser tool, now it is latest PS >> help >> erase background and bang it is done, converted to a smart object to combine to copy somewhere else, PS did a good job of keeping the legs even though they are black, I had other plans and wanted full legs which did not exist, so used eraser to remove legs, I have the model ready to go:
Next I wanted a scene, the image can be changed a certain amount, it was captured in a dark environment so best placing into similar, I decided on a street at night wet pavement, 10 years ago it would be looking at stock image sites for a scene, the issue being that getting the angles right can be difficult to put the image exactly at the right position as fool the eye (height of camera, angle of lens), AI is a godsend here. First I need some legs, so off to AI, asked "woman walking towards camera at night", this gave after a number of tries some legs I could easily cut out with back ground eraser, angle can be corrected to a certain extent with lens correction just to tweak the angle:
Now to create the background, over to leonardo.ai, and I used the Realtime canvas, this allowed me to start the scene by pasting the model and legs, I am after a background not final scene, now I did not capture the creation process at the time, so have recreated it as best can, AI being AI will create an infinite number of backgrounds, the prompt is:
"lady walking towards camera. At night time, wet pavement, lighted by street lights, photo realistic, dark stared black sky, bokeh, parked car in distance, neon sign"
This was a first stab by AI, notice how the angles are right in relation to the model and legs, the fake AI lady is not important, she will be replaced by the real model in photoshop by simply putting the model on top of the generated lady later on, here is first go:
Now here is where different images can be generated and refined, so, get one close to what is required, by changing creativity strength, and the prompt, then click output to input which replaces the input with the output, then the creativity strength can be changed (the higher the more random the image, do not go too far as would remove the fake lady), image copied back to input a number of times, change the input prompt to refine the image:
Once happy with the background we want a higher resolution to be used, so the Upscale image is clicked and the final background can be downloaded. Remnants of the fake lady can be removed with PS and the latest AI remove object to give a clean background. Paste the model and legs onto it, and do the usual PS processes (levels, saturation, contrast, smooth the skin on arms, body, parts of face, improve hair, a bit of smoothing, bit of dodging highlights, and burn mid tones to give more depth, do the same on face, arms, plenty of tutorials on improving models, took me around 60 mins, the modern un-sharpen mask recovered the original photograph well, I think AI is used with PS there also now as CS4 could not do what is done today). the final result:
AI is not the enemy, it is a creative tool.
Edited by -sp●●n-
Edited by -sp●●n-
