[Help] Can't generate specific seat type for a motorcycle image
Hi all,
I've been trying for over a week now to use Perchance Text to Image generator to generate the correct type of seat for a 1966 Harley-Davidson Electra Glide motorcycle image I'm working on. Here's a video of the type of seat I want the AI to use for my motorcycle image:
https://www.youtube.com/watch?v=UfXomqqQcYU
This is my best attempt at a descriptive prompt so far:
"A well preserved, vintage, 1966 Harley Electra Glide motorcycle in two tone red and white livery, with a post mounted 37-64 OEM Harley Vintage Buddy Seat."
'37-64' is the part number of this type of seat I believe, or at least it came up for a number of auction listings matching the seat shape, profile and compatibility with the Electra Glide Shovelhead series. While this has been getting me more suspension post mounted seats that look closer to the 1960's era, I haven't been able to get the seat in the video yet.
If anyone figures out a way to get what I want, please tell me how you did it.
Cheers everybody.
Hi, I am new to AI but I use google to find the actual names of the things I want the AI to generate, in you case try incorporating "sprung, solo vintage police motorcycle seat" into your prompt. that is what type of seat that is.
Thankyou for your comment and suggestion,
I gave your idea a try, a few different ways actually, but unfortunately I haven't been able to get consistent seats that look anything like the video example. The AI remains extremely wild in its RNG and the more I play around with this, the more I'm convinced the AI has no seat recognition ability to pick up on any keywords I give it.
If you or anyone else has further suggestions or ideas, please share them and I will try again.
Hi
That unpredictability factor is what is really screwing with me also, I like repeatable results, and the ability to tweak the item but that does not seam to be a feature of the Free AI Image Generators, or at least I just have not figured out how to yet.
The only other thing I can think of is prompt picture referencing, and I don't know if the Perchance Free text to AI Image Generator has this ability. a lot of the non free AI Image Generators have that ability.
Just out of curiosity , what kind of Image are you trying to create?
It's more for a personal project right now, but I wouldn't mind sharing the image later with others. The image is designed to capture the spirit of 'freedom' and 'fun', which is why I chose the 1966-1978 Electra Glide with its ridiculous bigger seat parked in a barn of 18th or 19th century persuasion with countryside stretched out into the distance under a clear blue sky beyond the wide barn entrance.
I managed to get all of the elements of my image generating consistently ~9-12 out of 12 images except for the motorcycle seat. The seat is a total wild card, the AI wildly guesses randomly every image and doesn't give me the seat type I want no matter how much or well I describe its shape and where it connects/sits relative to the frame and other components.
It's like the AI doesn't know what the seat looks like thus it guesses randomly. The thnig is, it CAN generate the seat in that example video, the one I want, but I only got it to happen 1 time out of ~12200 images generated 15 or 16 days ago. I haven't managed to get another one since. It is possible, but the problem is it's pure chance or luck, and this is wasting too much time.
Hi
Using just this prompt "sprung, solo vintage police motorcycle, "two tone red and white livery seat". " and a snap shot from the video of the seat I was able to get this within 10 tries.
This is the snapshot I used With some tweaking to the prompt and a better image you should be able to recreate what you looking for.
Hi Have you tried other free AI text to image generators? Before switching to Perchance I used https://freegen.app/, it has limits on what it is allowed to do "like not showing naughty stuff" but it does seem to do a good job of recreating none human based items and things. You might want to give it a try if you haven't already. One thing I like about it is you can add "1" image prompt, so you might be able to in corporate the seat that way by having a picture reference.
I did try a few with image upload capability, but they couldn't get the seat correct either. I think the AI and tools need more time to develop, so I will try this again in a few years.
Hi
Here is a video you might find helpful. https://www.youtube.com/watch?v=7h87NOrkc1M