GaggiX

19 ms·

331.
by GaggiX 2y ago
Yeah their approach of using two different models to detect and then inpaint is very subliminal given that many watermarks are semi-transparent. They could have just trained a UNet with adversarial loss + LPIPS to do all the work and it wou
332.
by GaggiX 2y ago
As a consumer you care about speed tho, and GPT-4.5 is extremely slow, at this point just use a reasoning model if you want the best of the best.
333.
by GaggiX 2y ago
It's better on AIME '24, Multilingual MMLU, SWE-bench, Aider’s polyglot, MMMU, ComplexFuncBench while being much much cheaper and smaller.
334.
by GaggiX 2y ago
The only thing I think ChatGPT is better is native image generation, their model is able to work in the joint-space much better than any other model I have seen, but I'm sure Google will try to catch up rapidly.
335.
by GaggiX 2y ago
We don't know what will happen in three days at this point.
336.
by GaggiX 2y ago
Retrofuturism, using AI and not having the editor set up to render Unicode characters.
337.
by GaggiX 2y ago
I don't even think it will be newer studios and that's it, probably well known studios will adopt it after a production ready solution is presented and battle tested, I doubt many will complain about not making in between frames.
338.
by GaggiX 2y ago
Yeah it will be much better than the low budget 3D models in anime, hopefully there will be a production ready product that works at a high enough resolution and studios will probably adopt it instead of using cheap labor.
339.
by GaggiX 2y ago
This is just a diffusion text-to-image model like many others, completely different than a LLM with a native image support.
340.
by GaggiX 2y ago
GPT-4o is the only model that seems to work well in the text-image joint space to this degree, even Gemini Flash 2.0 with native image support is not nearly as good so it will probably be a while for a good open source alternative to pop up
341.
by GaggiX 2y ago
That blog post is a year old. There has been a lot of progress since then: https://doubiiu.github.io/projects/ToonCrafter/
342.
Fiction.LiveBench: The First Long Context Benchmark for Writers(fiction.live)
2 points by GaggiX 2y ago|0 comments
343.
by GaggiX 2y ago
Claude is much better than ChatGPT at low-resource languages, at least it was a year ago, I haven't tested on new models from OpenAI but I believe that Claude still has an edge. For example, when ChatGPT was outputting nonsense in Geor
344.
by GaggiX 2y ago
An uninteresting complaining article with only clichés like "Nobody asked for it and nobody wants it."
345.
by GaggiX 2y ago
The native image support of GPT-4o is really impressive, no model before was able to work so well in the joint space, just compare it to the Gemini 2.0 Flash model with native image support, the difference is massive, so I'm not too su
346.
by GaggiX 2y ago
I use them everyday and they work greatly, I even made a command (using Claude, actually Claude made everything in that script) that calls Gemini from the terminal so that I can ask for question related to the shell directly there, just doi
347.
by GaggiX 2y ago
I can remember for example when I was wrong and how and still responding correctly, I don't have to forget my wrong answer to respond with the correct one.
348.
by GaggiX 2y ago
Why "selectively forget" should be a piece for AGI?
349.
by GaggiX 2y ago
Yeah it's just metadata.
350.
by GaggiX 2y ago
Ideogram 2.0 and Recraft also create images that looks very much real. For drawings, NovelAI's models are way beyond the uncanny valley now.
351.
by GaggiX 2y ago
There is literally a project page dedicated.
352.
by GaggiX 2y ago
If you are using a quadcopter you are probably close enough to control the drone directly.
353.
by GaggiX 2y ago
Compare to YouTube it has almost no content, like what if I want to see a gameplay of Balatro.
354.
by GaggiX 2y ago
Youtube is the real tough one to replace, BiliBili is probably the only one that comes close to it and it's Chinese.
355.
by GaggiX 2y ago
Fun fact the emulator works fine on Ladybird too, even thought it's very slow now.
356.
by GaggiX 2y ago
Given the topic, does anyone has some cool homebrew ROM for the gameboy?
357.
by GaggiX 2y ago
He was talking about STT models, not TTS. Whisper is open source and a good solution in many cases (in particular finetuned ones).
358.
by GaggiX 2y ago
The AI anime girl has 6 fingers btw, combating AI bot with AI girls. Edit: I will probably send a pull request to fix it.
359.
by GaggiX 2y ago
Translation technology has gotten so much better in the meanwhile, I didn't even notice it at first.
360.
by GaggiX 2y ago
Go can have data races, so I would not consider it memory safe.

More ›