Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Stable Diffusion print-on-demand apparel

1sewn.com

Post reply on HN
Print view View original

41–50 of 67 posts

Re: Show HN: Stable Diffusion print-on-demand apparel

#41

Post by hahamrfunnyguy »

The idea of using stable diffusion to create a t-shirt is compelling, but the execution here leaves a lot to be desired.

Randomly, I picked the stencil style and went for "Man with beard base jumping from the Eiffel tower." It produced one really poor image. It wasn't stenciled looking at all.

Next I went for pop art and used the prompt "Frog riding a motorcycle". The art style looked like something that an 8 year old would draw, not Andy Warhol (which I believe is the artist in your screenshot).

So I think you have some work to do in taking the inputs from the user and crafting them into better SD prompts that generate outputs that look like t-shirt artwork and in the styles they're supposed to be. I'd also drop the voice prompt support, no added value here. Definitely would prefer to type it on my computer and mobile keyboards already have a voice-input button.

Re: Show HN: Stable Diffusion print-on-demand apparel

#42

Post by snovv_crash »

zwog wrote: post #25
pinum wrote: post #10

What's the reason for the "everywhere except Europe" shipping? I'm guessing some legal/regulatory thing, but would be curious to know the details.

As stated in the original post printing and shipping is handled by https://www.printful.com/ and they offer their services in Europe. I assume it has something to do with copyright regulations in Europe or similar?

GDPR maybe if the microphone data is saved?

Re: Show HN: Stable Diffusion print-on-demand apparel

#43

Post by systemicdanna »

Many complained about voice-only input. I want to highlight a few other aspects that could be improved:

1. Navigation doesn’t work. Sometimes swiping right goes to the next page, sometimes doesn’t. Then you are stuck on the 3rd page and can’t go back.

2. What are points? I have 5 points, got to the payment step - the T-shirt is $25. How am I supposed to use points?

3. The T-shirt that is displayed as a 3D background is completely different from the low quality generic fit design at the end. This discrepancy is off-putting.

4. The tiny final print is unintelligible on my phone. Some squiggles and circles that I can’t make sense of at this resolution/zoom level.

I think the idea of custom fashion is interesting, but this implementation doesn’t feel empowering at all. Feels limiting and gimmicky. More importantly, the result - a generic T-shirt with a very small print is not fashionable or even attractive.

Re: Show HN: Stable Diffusion print-on-demand apparel

#44

Post by pvg »

snovv_crash wrote: post #40
pvg wrote: post #37

Earlier quoted context omitted.

'Don't be a jerk' is the more important, seemingly less obvious alternative.

Requesting microphone permissions for something that would actually be easier without it is also very obnoxious, and for anywhere with GDPR perhaps even illegal if the data was persisted. Maybe this is why it isn't available in Europe?

The guideline is there to avoid escalating obnoxiousness because it trashes the forum.

Re: Show HN: Stable Diffusion print-on-demand apparel

#48

Post by mdaEyebot »

hahamrfunnyguy wrote: post #41

The idea of using stable diffusion to create a t-shirt is compelling, but the execution here leaves a lot to be desired. Randomly, I picked the stencil style and went for "Man with beard base jumping from the Eiffel tower." It produced one really poor image. It wasn't stenciled looking at all. Next I went for pop art and used the prompt "Frog riding a motorcycle". The art style looked like something that an 8 year ol…

It's easier to get good results by generating 2-8 images for each prompt, picking the best result, using that as the starting point for a new run.

Tweak the parameters of how close the new image should be to the old one and the prompts, tweak the prompts to address problem areas in the image, generate another 2-8. Rinse and repeat until you have something decent.

I doubt that the really impressive SD images are coming from submitting a prompt once and taking that output. It's better than past open-source efforts like the mini dall-es, but it also still has a way to go before it can reliably produce good results on its own.

AI art on products is an interesting idea, but personally I would use a more fully-featued SD web UI to generate the image locally.

Re: Show HN: Stable Diffusion print-on-demand apparel

#50

Post by omarfarooq »

AkshatJ27 wrote: post #9

I expected it to somehow fill the entire shirt, but apparently it just makes a rectangular patch of the image on a plain black/white shirt. Also, the speech to text doesn't seem to do that well.

That would require the more involved 'cut and sew' process.
Post reply on HN