Help Blending Images to make better composite

[SolidSnake2003]

I have another version of the image I have been working on for a fanfiction project I am working on based on Stephen King's IT story. The version I worked on before had alot of shadows and other elements that are harder to hide. This is a more simple replacement without the complex shadows and shapes.

This is the base image: https://i.imgur.com/ibFKBem.jpg This is from an episode of The Good Doctor. I am wanting to replace characters from the show with the characters from IT.

This is what I have so far: https://i.imgur.com/by90NN3.jpg

I have matched the new face of the mother pretty well. I have matched the color of the boy's head as best as I could. But the biggest obstacle for me is making it blend seamlessly. This is the base image of the boy's head: https://i.imgur.com/7ldZFf1.jpg This is the best image I could find of the actor that was in a close enough angle.

[SolidSnake2003]

Would this head image work better?

https://i.imgur.com/1MdTeap.png

[SolidSnake2003]

Here is the project file of what I have done with the layers in place and the coloring effects to make them match as best as possible.

 

chap02_georgie-sharon_hospital.afphoto

[firstdefence]

The eye and brain are very good at recognising lighting anomalies, so, the shading is wrong on Georgies head you instinctively know this is wrong because it doesn't look right in the environment it's placed. The left side of the face as you look at it should not have shadow and the lower right part of the face should have a gradient shadow. The hair of Georgie looks like a helmet so the straight edges needs ragging up a bit using a small texture brush set to a small size 50% hardness and rotation at 100% under the brushes dynamic settings.

This helmet effect is not good

This looks a bit more natural

The other stand out is the resolution of the Georgie image, it's poor quality so looks soft and fuzzy, the tonality is also off, looking flat and a bit washed out and dirty, the shadow created by the head is too big.

The bear is also flat and obviously distorted, once an image is distorted it looks out of sorts, and its very difficult to get an invisible distortion because, again, the eye is very good at subconsciously marrying up fine detail to it's environment. the bear looks dull and washed out and dirty in comparison to the environment so it will always look out of place.

 

The mum face needs to have a basic eraser brush set to zero hardness and brushed over the edges as a hard line can still be seen when zoomed in.  The complexions are different with the base image having a more flushed pinkish complexion while the face overlaid has a green/yellow tone to it. 

A slight adjustment to the Curve layer on the green curve will bring the overlaid face into a closer colour match, but the the middle might need a bit more work. Once again the image is soft compared to the base image so the texture is mismatched which make the image look out of sorts.

 

[SolidSnake2003]

25 minutes ago, firstdefence said:

The eye and brain are very good at recognising lighting anomalies, so, the shading is wrong on Georgies head you instinctively know this is wrong because it doesn't look right in the environment it's placed. The left side of the face as you look at it should not have shadow and the lower right part of the face should have a gradient shadow. The hair of Georgie looks like a helmet so the straight edges needs ragging up a bit using a small texture brush set to a small size 50% hardness and rotation at 100% under the brushes dynamic settings.

This helmet effect is not good

This looks a bit more natural

The other stand out is the resolution of the Georgie image, it's poor quality so looks soft and fuzzy, the tonality is also off, looking flat and a bit washed out and dirty, the shadow created by the head is too big.

The bear is also flat and obviously distorted, once an image is distorted it looks out of sorts, and its very difficult to get an invisible distortion because, again, the eye is very good at subconsciously marrying up fine detail to it's environment. the bear looks dull and washed out and dirty in comparison to the environment so it will always look out of place.

 

The mum face needs to have a basic brush set to zero hardness and brushed over the edges as a hard line can still be seen when zoomed in.  The complexions are different with the base image having a more flushed pinkish complexion while the face overlaid has a green/yellow tone to it.

 

Thanks for the advice, I basically just started this one, haven't done as much with this one as I did the other one with more shadows. I was mainly just trying to get things in place and fine tune it to blend everything together after getting everything lined up. I didn't add any shadow underneath Georgie's head, it might be the original head showing showing through The coloring I knew was off especially with Georgie's head. I tried to match as best as I could with color balancing, and selective color. I also used a 50% gray color layer as a check layer to match the colors as best as I could.'

 

Here is a screen capture I took of Jackson Robert Scott from another movie he was in, but I don't know if I could manipulate this head to match with the original: https://i.imgur.com/1MdTeap.png

[SolidSnake2003]

7 hours ago, firstdefence said:

The eye and brain are very good at recognising lighting anomalies, so, the shading is wrong on Georgies head you instinctively know this is wrong because it doesn't look right in the environment it's placed. The left side of the face as you look at it should not have shadow and the lower right part of the face should have a gradient shadow. The hair of Georgie looks like a helmet so the straight edges needs ragging up a bit using a small texture brush set to a small size 50% hardness and rotation at 100% under the brushes dynamic settings.

This helmet effect is not good

This looks a bit more natural

The other stand out is the resolution of the Georgie image, it's poor quality so looks soft and fuzzy, the tonality is also off, looking flat and a bit washed out and dirty, the shadow created by the head is too big.

The bear is also flat and obviously distorted, once an image is distorted it looks out of sorts, and its very difficult to get an invisible distortion because, again, the eye is very good at subconsciously marrying up fine detail to it's environment. the bear looks dull and washed out and dirty in comparison to the environment so it will always look out of place.

 

The mum face needs to have a basic eraser brush set to zero hardness and brushed over the edges as a hard line can still be seen when zoomed in.  The complexions are different with the base image having a more flushed pinkish complexion while the face overlaid has a green/yellow tone to it. 

A slight adjustment to the Curve layer on the green curve will bring the overlaid face into a closer colour match, but the the middle might need a bit more work. Once again the image is soft compared to the base image so the texture is mismatched which make the image look out of sorts.

 

Don't know if you saw this earlier topic of mine, but here is the other variation I've started using a screen capture as the base instead of a production still: https://i.imgur.com/gnNPQW6.jpg
 

[firstdefence]

5 hours ago, SolidSnake2003 said:

Don't know if you saw this earlier topic of mine, but here is the other variation I've started using a screen capture as the base instead of a production still: https://i.imgur.com/gnNPQW6.jpg
 

Yes I saw that, I think the scenario is similar to this one, ultimately, if the lighting of the subject/object doesn't match the scene it's going to be difficult to make it fit, the more you edit the replacement image in the scene the less likely it is to look realistic, when you start to try and distort the image to fit a perspective that makes it worse, in that scenario the liquify tool is probably the best option but used in small amounts.

While both projects are good exercise's I doubt you would get a decent enough result unless you were fortunate enough to find a clip that very closely matched the lighting, was a decent resolution, was close in colour and tone and the head position was very close. 

[SolidSnake2003]

3 hours ago, firstdefence said:

Yes I saw that, I think the scenario is similar to this one, ultimately, if the lighting of the subject/object doesn't match the scene it's going to be difficult to make it fit, the more you edit the replacement image in the scene the less likely it is to look realistic, when you start to try and distort the image to fit a perspective that makes it worse, in that scenario the liquify tool is probably the best option but used in small amounts.

While both projects are good exercise's I doubt you would get a decent enough result unless you were fortunate enough to find a clip that very closely matched the lighting, was a decent resolution, was close in colour and tone and the head position was very close. 

Yeah I know that it works best when you work with material that already closely matches. But unfortunately that is not always the case to get scenes to work with for certain actors. The only role that Jackson Robert Scott has where he goes to the hospital is a movie called The Prodigy. Here is a website with a few screencaps from the scene in the movie: https://www.teenidols4you.com/picture.php?pg=2&pe=jackson-robert-scott&act=3360&mv=5881 I have also found a couple of stock images where the coloring was not too different from an image from The Prodigy, but I can't afford any of the pricing for the stock site. I know that Shutterstock has like a free 7 day trial or so with 10 images during that time

[v_kyr]

46 minutes ago, SolidSnake2003 said:

... I know that it works best when you work with material that already closely matches ...

Hint: make also use on the net of find similar picture services, so things like ...

• https://lens.google.com
• https://images.google.com/
• https://tineye.com/
• ... etc.

[SolidSnake2003]

22 minutes ago, v_kyr said:

Hint: make use on the net of find similar picture services, so also things like ...

• https://lens.google.com

I remember Google Image search having a find similar image function once, but can't seem to find it. 

[v_kyr]

7 minutes ago, SolidSnake2003 said:

I remember Google Image search having a find similar image function once, but can't seem to find it.

You can also search on the net after help for that ...

• using google images
• using google lens
• ...

[SolidSnake2003]

6 minutes ago, v_kyr said:

You can also search on the net after help for that ...

• using google images
• using google lens
• ...

Does Lens work on PCs cause I don't use andriod phones for photo editing.

[v_kyr]

4 minutes ago, SolidSnake2003 said:

Does Lens work on PCs cause I don't use andriod phones for photo editing.

You can use that in/from whatever webbrowser, I use it sometimes under MacOS via Firefox/Chrome ...! - Just try out some of the links I showed you above.

[SolidSnake2003]

Here I've tried the image search using an image I through google:

[SolidSnake2003]

I guess it works alot better when you use a bunch of images, and not just work from one base image.

[firstdefence]

Maybe there is an opportunity to use one of the Ai sites such as:

https://playgroundai.com

https://www.midjourney.com/

https://pebblely.com/

https://magicstudio.com/

https://open.ai/ (restricts what can be created, so, boy in hospital bed will be seen as inappropriate, but male in hospital bed will not.)

to name a few and to generate an image that fits the base image, this might take a bit of learning but maybe an avenue to explore, there are quite a few books on how to use these sites and the prompt feature. https://learnprompting.org/

[SolidSnake2003]

9 minutes ago, firstdefence said:

Maybe there is an opportunity to use one of the Ai sites such as:

https://playgroundai.com

https://www.midjourney.com/

https://pebblely.com/

https://magicstudio.com/

to name a few and to generate an image that fits the base image, this might take a bit of learning but maybe an avenue to explore, there are quite a few books on how to use these sites and the prompt feature. https://learnprompting.org/

Doesn't it cost money to use AI?

[firstdefence]

Try Open ai it's free to use, some do cost money but many are free, it would be worth searching for free ai image generators, it takes a bit of practice but once you get your head around it you can get reasonable results.

[firstdefence]

This will help fine tune images: https://learnprompting.org/docs/category/️-image-prompting

[SolidSnake2003]

24 minutes ago, firstdefence said:

This will help fine tune images: https://learnprompting.org/docs/category/️-image-prompting

This is what I should learn how to do? Learn AI prompting, then super impose Jackson's or others faces onto these generated images?

[thomaso]

7 hours ago, firstdefence said:

prompt feature. https://learnprompting.org/

Nice source! I would not have thought that multiple repetition of "very" in combination with "beautiful" would have this affect on waterfall pictures.

But whatsapp here: Smart reliability, offering various languages without using AI ?
[ btw: is "Bild" English, too, and wherefrom comes "promting"? ]

[firstdefence]

4 hours ago, thomaso said:

Nice source! I would not have thought that multiple repetition of "very" in combination with "beautiful" would have this affect on waterfall pictures.

that's very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, very, surprising lol! makes you wonder how many very's would be enough 

[SolidSnake2003]

I find this AI stuff confusing