Are we getting built-in AI render in Chief anytime soon? Softplan has it already.

[ComputerMaster86]

I must of also tried the following prompt at some point...

 

"Create a HDRI photo realistic rendering. Keep integrity of original image. Do not change any image details. Do not move or modify features. Keep all colors the same. Add natural sunlight. Keep original image Size."

 

[basketballman]

5 hours ago, ComputerMaster86 said:

Here is an image I exported from Chief and tried in chat GPT a few months back...

 

 

Here was the output from chat GPT...

 

 

It added a shingle ridge cap to a steel roof.    It misinterpreted the porch beam and that was a little off.

 

I used "Create photo realistic rendered image. Keep integrity of original image."  I probably needed to use a better prompt.

 

   Yes, negative prompts need to be very specific such as: 

DO NOT CHANGE OR REVISE LAYOUT-WALL CONFIGURATION, DOOR, WINDOW SIZES AND LOCATIONS OR ANYTHING ELSE.

Do not change camera view, angle, field of view, position, distance or aspect ratio.

 

 

[Rich_Winsor]

Here's a thought. Has anybody tried creating

one image using a prompt in lower case and

then run the exact same prompt in upper case

to see if there is any difference in the images if

you yell at AI. 

 

 

[basketballman]

"Yelling doesn't work; I've tried it many times.." 

[Renerabbitt]

So many of these problems would be mitigated by learning some very basic beginner photoshop combined with Krea Legacy. I used NO prompts, none whatsoever, i didnt even have to think about prompting. This edit, albeit not my best, took 10 minutes
 

Source Image

Krea Output:

Masked Area for errors in the AI, things that the AI Changed

Then add a Camera Raw Filter

[basketballman]

But who wants to learn P.S., Krea Legacy & how to place a Camera Raw Filter and then spend 10 minutes waiting ...

            when the right prompt and 30 seconds results in this:

  And the lighting is so much better...

[Renerabbitt]

2 hours ago, basketballman said:

But who wants to learn P.S., Krea Legacy & how to place a Camera Raw Filter and then spend 10 minutes waiting ...

            when the right prompt and 30 seconds results in this:

  And the lighting is so much better...

I don't think one can call themselves a rendering artist if they don't learn at least basic photo editing. You certainly couldn't get into commercial rendering work with FF&E markets without knowing how to mask.
I completely disagree that the lighting is better, your image depicts hard sun lighting with a skyscraper shadow. Things are overexposed and details blown out. A stock home site would reject this image.(I do a lot of stock home image edits)
ALSO
I did NOT want the atmosphere of the shot to change...one of the biggest pitfalls of Ai
You and I my friend can argue to the cows come home. I did not want the sun rotated, or a skyscraper casting a shadow, I don't find that this was that great of an enhancement, I did not want my image stretched, and we both know that ai will fail on a more complicated house in many different ways. I don't find this image to be in the same category of LOD.
The grass, roofs, and battens do not look improved. The backdrop sky is underexposed and looks like a pasted image.
What I DID want, is a true pixel to pixel perspective improvement. Your AI changed the perspective and skewed the image making it ver complicated to material ID or Object ID map(rendering technician must-haves)
You have responded with similar posts before where the Ai changed things. Its far faster to mask them out then go back and forth with Ai prompting
 

Edited 7 hours ago by Renerabbitt

[basketballman]

30 minutes ago, Renerabbitt said:

I don't think one can call themselves a rendering artist if they don't learn at least basic photo editing.
You and I my friend can argue to the cows come home. I did not want the sun rotated, I don't find that this was that great of an enhancement, I did not want my image stretched, and we both know that ai will fail on a more complicated house in many different ways.
What I DID want, is a true pixel to pixel perspective improvement. Your AI changed the perspective and skewed the image making it ver complicated to material ID or Object ID map(rendering technician must-haves)
You have responded with similar posts before where the Ai changed things. Its far faster to mask them out then go back and forth with Ai prompting
Try this one on for size:

Its a rendering from Chief with no post editing. I know what chatgpt will do to it already, masking is the only solution

  For me, the small shift in camera position is well worth the trade off in enhanced quality & sharpness.

  And except for a small camera shift I don't see much that has changed from the original, except for the better.

[Renerabbitt]

1 hour ago, basketballman said:

except for the better.

I am sorry but I disagree. All this did was add a basic filter. Over saturated and black and highlight levels too high.
Shift in camera position wont work for many use cases where we want a material ID or Object ID mask, it has to be pixel to pixel, and that would be one of the differences between a consumer based approach to professional. All of the big firms ask you to provide the mask in a PSD file so things like floor or cabinet colors can be changed by adding hue and saturation masks. This makes it very easy to say, change the material on the floor.
This was 10 min in photoshop. I don't think you will agree that it is improved but this is closer to what the client wanted

Edited 7 hours ago by Renerabbitt

[SHCanada2]

the sales/marketing people I deal with want blue sky and sun. basketballmans has too much exposure where the sun washes out (and why the odd shadow into the street?), but rene, they would think yours is too grey.

3 hours ago, Renerabbitt said:

 

But I'm in Alberta which has the most sun in all of Canada, so maybe they are biased. grey sky here =depressing

 

[Renerabbitt]

57 minutes ago, SHCanada2 said:

the sales/marketing people I deal with want blue sky and sun. basketballmans has too much exposure where the sun washes out (and why the odd shadow into the street?), but rene, they would think yours is too grey.

 

But I'm in Alberta which has the most sun in all of Canada, so maybe they are biased. grey sky here =depressing

 

That was mixing up the points that I was trying to make which is that this image was not for a stock plan site it was to edit and improve an existing rendering which was the call intent of that original post but then I did want to illustrate that basketball man's rendering drastically change the environment. I don't really understand why I feel compelled to fight for what I was trying to explain in a pretty simple and straightforward post which is masking out a problem area of an AI-based rendering would be the most basic and substantial improvement to any rendering and then on top of that at one of the most important aspects of providing rendering services is keeping all iterations of an image perspectively true so that new rounds of edits can be made without affecting old routes of edits. Forgive any lack of punctuation this was voice to text

[GaryVanScyoc]

I do believe in time Chief will have some form of AI 'built in' out of necessity. In the meantime I've been kicking the tires of Archsynth. My first impressions: WOW!! But, along with the wow, there is a certain amount of, "now why did it have to do that"! It didn't take long to figure out, bit of learning curve here to be able to keep Archsynth from misbehaving. An example of the 'wow' factor below. I uploaded a line drawing of a house I'm developing in Chief Architect. I gave Archsynth two prompts: Set in a forest and cloudy day. In 30 seconds I got the result as shown.