maalber

8 ms·

61.
by maalber 2y ago
Hey Viktor, looks cool! It looks like it uses some image-to-video generation model. I would be interested to know which one you used, how you selected it, and whether you did some training/fine-tuning yourself in the process.
62.
by maalber 2y ago
Looks cool, although i personally stopped at the login prompt. I would probably also tone down the continuous generic pop-up that someone generated something. It seems a bit too frequent, repeating and generic to be trustworthy. That being
63.
by maalber 2y ago
This is actually really cool. First tried with one of the example images which of course worked really well. I then tried with a couple of images I had lying around on my own PC which also worked surprisingly well! Nice work. A couple of q
64.
by maalber 2y ago
Interesting, what was the inspiration for this? :D Also, I discovered that it is also possible to manipulate it with mouse as well as the arrow keys
65.
by maalber 2y ago
Seems like a relatively niche, but potentially helpful tool. I might try it out properly later. I have a couple of questions: There seems to be a fair bit of effort put into this, and you are seemingly offering it completely for free. Do yo
66.
by maalber 2y ago
Cool, and quite intuitive UX. At first I did not understand what to do, but after the first round and getting one of the categories correct, it was clear how it works. I am not familiar enough with birds to properly play it, but maybe I w
67.
by maalber 2y ago
The version we currently have open is only 2, but we are soon going to release a version that is more general ranking which can do 3 and more options. Hit me up and I can keep you updated when it is available!
68.
by maalber 2y ago
Adding to this, ELO is also heavily used when evaluating, benchmarking and ranking state-of-the-art generative AI models. Particularly LLMs, image generation and video generation models.
69.
by maalber 2y ago
Yes, I also feel like most of the off-the-shelf models struggle with generating nice and natural looking logos, especially something that resembles vector graphics. Since you have been fine-tuning the models yourself, I really think the too
70.
by maalber 2y ago
Very interesting - definitely looks better than what i mostly get from DALL-E Can you share some technical details on how you did it? Did you train your own model? Or fine-tune an existing? The option to polish it in figma is interesting. I
71.
by maalber 2y ago
This is so funny lol! Have you considered donating (parts) of the charged amount to charities?
72.
by maalber 2y ago
This is great, and I love your presentation of it - hilarious! "On average it costs about 10 to 30 cents per search. It's also extremely slow."
73.
by maalber 2y ago
Nice, quick work!
74.
by maalber 2y ago
Cool! Can you elaborate a bit on the technology? Are you using off the shelf models or did you train your own for the purpose?
75.
by maalber 2y ago
I tried it out quickly - looks interesting. Can you elaborate on what the idea/intention/plan for this is? It seems you have to be careful with the fire :D I thought I would be able to put it out with the water or vapour, but that
76.
by maalber 2y ago
Interesting, that is a good size userbase! And good argument regarding the ranking vs rating - did not consider this. I have a site, app.rapidata.io/compare which has a related feature, allowing you to have 100 people vote between two
77.
by maalber 2y ago
Very interesting work. I am curious as to whether you can share some more of your technical process. I guess you are using some pre-trained diffusion model that you have somehow altered and fine-tuned? How did you decide on training data?
78.
by maalber 2y ago
Very interesting. I am curious, how big is your community, and how did you advertise it to grow it to the point it is at currently? And how many are active, i.e., how long does it normally take before your images have been ranked? This remi
79.
by maalber 2y ago
Interesting idea! It is kind of like an advanced version of hangman. I ran into a couple of user experience issues where it was not clear to me if I had guessed the letter, as the box was not checked. Also the exclusion of non-possible let
80.
by maalber 2y ago
Cool! works quite well. The images on your landing page seem a bit stretched though
81.
by maalber 2y ago
Came to the comments with the same question ^^ I thought it was a tool to help downvote bots found in the wild
82.
by maalber 2y ago
Cool, can you share some detail on how this was made? Does the flipping actually have any effect?
83.
by maalber 2y ago
Very cool! I like the simplistic design. Would it make sense to have a short time limit on the answer to avoid people quickly popping over to wikipedia?
84.
by maalber 2y ago
I have made a Figma plugin that quickly lets you get human feedback and preference between two designs. Select two frames and launch the plugin. Then describe the criteria that you want the two options to be compared based on, and launch to
85.
Show HN: I created a Figma plugin for small AB tests with real people in minutes(figma.com)
2 points by maalber 2y ago|1 comments
86.
Ranking of popular image generation AI models (incl. Flux) from 2M votes(arxiv.org)
1 points by maalber 2y ago|1 comments
87.
by maalber 2y ago
Efficiently evaluating the performance of text-to-image models is difficult as it inherently requires subjective judgment and human preference, making it hard to compare different models and ultimately quantify progress. We developed a syst
88.
by maalber 2y ago
This is a really cool idea! I am also interested in a bit more info about the technology and methods behind. Is it using off the shelf LLMs and system prompts to generate code based on the request? Or is it a specifically trained model for
89.
by maalber 2y ago
Can you elaborate a bit on the technical details? Do you train you own models or integrate with APIs from e.g. OpenAI? If you train your own models, what kind of challenges did you face regarding low-resource languages (smaller languages).
90.
by maalber 2y ago
Understood! I do not have an easy way to access this in the browser currently. I will consider adding this in the future!

More ›