News for hackers. Stuff that matters.
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by tqkxzugoaupvwqr on Monday November 14, 2022 @11:36PM (#33605336)
This is a copy cat of https://avatarai.me/ and https://www.profilepicture.ai/ which launched first and published their approach and revenue. Both are wrappers around https://www.astria.ai/ afaik.
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by noduerme on Monday November 14, 2022 @11:58PM (#33605479)
They're just wrapping and reselling the Astria API? That seems dangerous given the potential for their own customers to abuse the method for deepfakes.
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by alex-polymath on Tuesday November 15, 2022 @06:29AM (#33607892)
No, we training dedicated neural network for each user based on stable diffustion 1.4. That's why it takes so long
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by throwoutway on Tuesday November 15, 2022 @08:19AM (#33608778)
What do you mean? If I upload an image of myself to the Astro’s API, it can be used ins deep fake?
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by rahul_nyc on Monday November 14, 2022 @11:38PM (#33605361)
I've been working on Picasa AI[1] for sometime, we use cloud GPUs for building the Stable Diffusion + Dreambooth models. I'm wondering what you folks are using? Any API providers?

[1] - https://PicasaAI.com (Founder here)

Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by nagonago on Monday November 14, 2022 @11:29PM (#33605292)
Cool idea, although the pricing seems pretty steep to me. Knowing the internet, similar services will likely be available for free or at least cheaper soon.

Also, this should probably be a Show HN instead.

Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by ronsor on Monday November 14, 2022 @11:32PM (#33605318)
I'm slightly reminded of the "$400 for Dreambooth" post on the Stable Diffusion subreddit: https://i.redd.it/byu74ix1itz91.png

Cheaper services will no doubt pop up next week

Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by synapticpaint on Tuesday November 15, 2022 @07:58AM (#33608599)
Cheaper services are here now: https://synapticpaint.com/ I charge $2-$4 per model training.

I don't think that the profile pic/avatar space is going to be very profitable. As you can see from this post, there is quite a bit of competition already. Because this is not that hard to set up, there is nothing stopping someone from charging the cost of production for this, which is essentially what I'm doing (I added in a small amount of overhead as a buffer for non-GPU compute and in case I messed up my math, math is hard). That's why I'm focused on building for other use cases instead (making content creation more accessible).

Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by alex-polymath on Monday November 14, 2022 @10:47PM (#33605026)
I created 200 avatars for me and my sister using stable diffusion 1.4 fine-tuning and wrapped it into simple service - https://portret.ai/
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by diimdeep on Tuesday November 15, 2022 @12:54AM (#33605796)
Why projects you submitted before are all offline ?
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by alex-polymath on Tuesday November 15, 2022 @01:03AM (#33605849)
One is alive - http://colorize.cc/
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by alex-polymath on Tuesday November 15, 2022 @01:04AM (#33605855)
@levelsio created like 70 indie projects. And only 3-4 are alive
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by boolean on Monday November 14, 2022 @11:11PM (#33605169)
I just tested https://astria.ai and impressed with the results. Curious how this is different.
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by vimy on Monday November 14, 2022 @11:10PM (#33605159)
What are you using for backend?

Typo: Is it one time payment? Yes, absolutely. No stupid subscriptinos.

Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by antisthenes on Monday November 14, 2022 @11:05PM (#33605135)
Is this a Show HN?
Re:Stable Diffusion 1.4 fine-tuning on real photos for makin... by ec109685 on Monday November 14, 2022 @11:33PM (#33605324)
It and many of these comments are ads.