Inframan
Established Member
- Impact
- 551
I'm writing this as a participation post of NP user experiences. I posted previously about how to get greater functionality out of the ATOM AI image creator. Since that post they upgraded the AI so it's a brand new ballgame. If anyone has a AI prompt they're willing to share 
.
My ChatGPT (image) usage, prior to the mandatory (sign in), was my primary AI creation tool. I exclusively used it to create images for DAN.
My satisfaction level was around 50-70%. The more I used it the more the limitations of the AI were apparent. What's advertised and what it's actually capable of are (fill in the blank).
The main thing I noticed was, it absolutely would not spell a word correctly. Drawing a realistic human profile below the shoulders was rocket science. Having ChatGPT make snarky responses was also a negative. (Telling me it's time to change the topic.) ChatGPT can satisfactory draw nature scenes, cartoon imagery, illustrations (New Yorker Magazine).
I guess the hype factor of GPT is it's ability to combine or draw any art style conceivable, BUT is that practical?
I typed the most extreme ends of the artistic spectrum just to see. It's interesting in itself, but I didn't use it for any domains.
---------------
AI Photo Generator is my new testing ground. It's more robust than GPT. I might have tested half of the available art styles. It's ability to spell words is not accurate. It does have some ability to spell, however limited. I used it for more intricate artwork on DAN and before SquadHelp/ATOM had it's own AI tool.
When you choose photorealistic you're in the danger zone. The extra deformed limbs and the demon sunken eyes is disturbing (cursed photos). When you choose less than photorealistic your results are better.
I mostly used it for generic abstract background patterns for ATOM (before the AI).
My satisfaction level is 65-85%.
-------------------
ATOM's AI creator
It's not Free. Out of the three I've tested it's ability to spell is the best. It's accuracy is around 85-91%. This is not an ALL Inclusive regenerative AI. It's primary function is logo creation. As I posted before about this AI, it takes the domain name literally. The less visually descriptive your domain is , the more generic your image is.
I used the 50 free tokens and bought 100 more. It took an average of 8-10 attempts to get the image I wanted for each domain.
Why so many attempts? Because of my initial observations (with my own domains).
Left Side 2 tone logo and domain name.
Center aligned 2 tone logo above domain name.
Using the same font for different domain names.
Generic abstract logo. (this is more a limitation on the domain since the AI is very literal)
Since my upgrade to Sapphire, I examined over 4000 logo designs in (Tech, Internet, Software)
The lightning bolt design is the same for everyone. It's the same design as NP's (new posts) logo. 100's have this including myself.
The rocket ship design. Me 2
and the rest of the minions. Maybe it's appropriate that we all have the same rocket design.
The panda face design. I thought it was unique until I saw it several times. Myself included.
The bicycle design. I spent 10 tokens perfecting the image just to see the same bike design just a different color several times.
The farm / tractor / sunset There is a degree of variation but not by much. Not me, but still.
The Brain circuit board Me and 100 other sellers have this same design.
The Green Energy / Ecology / plant life We all have the same shade of green and basically a leaf theme. By far the greatest degree of repetition/generic design.
I will say 0.5-1.0% of all the domain logos I viewed were highly exceptional. I have no idea how they were prompted and it wouldn't make a difference because every domain is a snowflake.
My conclusion is the more Visually descriptive your domain name is the better the AI performs.
Feel free to share your observations.
.My ChatGPT (image) usage, prior to the mandatory (sign in), was my primary AI creation tool. I exclusively used it to create images for DAN.
My satisfaction level was around 50-70%. The more I used it the more the limitations of the AI were apparent. What's advertised and what it's actually capable of are (fill in the blank).
The main thing I noticed was, it absolutely would not spell a word correctly. Drawing a realistic human profile below the shoulders was rocket science. Having ChatGPT make snarky responses was also a negative. (Telling me it's time to change the topic.) ChatGPT can satisfactory draw nature scenes, cartoon imagery, illustrations (New Yorker Magazine).
I guess the hype factor of GPT is it's ability to combine or draw any art style conceivable, BUT is that practical?
I typed the most extreme ends of the artistic spectrum just to see. It's interesting in itself, but I didn't use it for any domains.
---------------
AI Photo Generator is my new testing ground. It's more robust than GPT. I might have tested half of the available art styles. It's ability to spell words is not accurate. It does have some ability to spell, however limited. I used it for more intricate artwork on DAN and before SquadHelp/ATOM had it's own AI tool.
When you choose photorealistic you're in the danger zone. The extra deformed limbs and the demon sunken eyes is disturbing (cursed photos). When you choose less than photorealistic your results are better.
I mostly used it for generic abstract background patterns for ATOM (before the AI).
My satisfaction level is 65-85%.
-------------------
ATOM's AI creator
It's not Free. Out of the three I've tested it's ability to spell is the best. It's accuracy is around 85-91%. This is not an ALL Inclusive regenerative AI. It's primary function is logo creation. As I posted before about this AI, it takes the domain name literally. The less visually descriptive your domain is , the more generic your image is.
I used the 50 free tokens and bought 100 more. It took an average of 8-10 attempts to get the image I wanted for each domain.
Why so many attempts? Because of my initial observations (with my own domains).
Left Side 2 tone logo and domain name.
Center aligned 2 tone logo above domain name.
Using the same font for different domain names.
Generic abstract logo. (this is more a limitation on the domain since the AI is very literal)
Since my upgrade to Sapphire, I examined over 4000 logo designs in (Tech, Internet, Software)
The lightning bolt design is the same for everyone. It's the same design as NP's (new posts) logo. 100's have this including myself.
The rocket ship design. Me 2
and the rest of the minions. Maybe it's appropriate that we all have the same rocket design.The panda face design. I thought it was unique until I saw it several times. Myself included.
The bicycle design. I spent 10 tokens perfecting the image just to see the same bike design just a different color several times.
The farm / tractor / sunset There is a degree of variation but not by much. Not me, but still.
The Brain circuit board Me and 100 other sellers have this same design.
The Green Energy / Ecology / plant life We all have the same shade of green and basically a leaf theme. By far the greatest degree of repetition/generic design.
I will say 0.5-1.0% of all the domain logos I viewed were highly exceptional. I have no idea how they were prompted and it wouldn't make a difference because every domain is a snowflake.
My conclusion is the more Visually descriptive your domain name is the better the AI performs.
Feel free to share your observations.
