glide-text2im VS pixray

GLIDE(2021) https://syncedreview.com/2021/12/24/deepmind-podracer-tpu-based-rl-frameworks-deliver-exceptional-performance-at-low-cost-173/ ||  소스코드 : https://github.com/openai/glide-text2im

The bolded part is a reference to the publicly released version of OpenAI's GLIDE, which is the predecessor of DALL-E 2. OpenAI didn't release the GLIDE model(s) trained on human faces.

OpenAI's GLIDE text2im https://github.com/openai/glide-text2im

If I can stop wasting my time playing video games for a while, I might work on getting the Dalle-2 open-source predecessor (GLIDE) to work. Also can't wait for this to be released, I have so many uses for it!

A few comments by someone who's spent way too much time in the AI-generated space:

* I recommend reading the System Card that came with it because it's very through: https://github.com/openai/dalle-2-preview/blob/main/system-c...

* Unlike GPT-3, my read of this announcement is that OpenAI does not intend to commercialize it, and that access to the waitlist is indeed more for testing its limits (and as noted, commercializing it would make it much more likely lead to interesting legal precedent). Per the docs, access is very explicitly limited: (https://github.com/openai/dalle-2-preview/blob/main/system-c... )

* A few months ago, OpenAI released GLIDE ( https://github.com/openai/glide-text2im ) which uses a similar approach to AI image generation, but suspiciously never received a fun blog post like this one. The reason for that in retrospect may be "because we made it obsolete."

* The images in the announcement are still cherry-picked, which is therefore a good reason why they tested DALL-E 1 vs. DALL-E 2 presumably on non-cherrypicked images.

* Cherry-picking is relevant because AI image generation is still slow unless you do real shenanigans that likely compromise image quality, although OpenAI has likely a better infra to handle large models as they have demonstrated with GPT-3.

The flags were generated using Glide. You can try it out yourself in Google Colab

Not sure I ran https://github.com/pixray/pixray, which I think is based off of VQGAN. It would fluctuate from 50-100 -> 300-400 watts every few seconds and my desk lamp on the same power strip would flicker dim every time it hit. I game all the time at that wattage, so I assume it was spiking well over 400 when it would kick up.

I used code from https://github.com/pixray/pixray. It is not easy to use directly, and you need datacenter GPU to use pixray well. I ended up maintaining the texts at google sheets and running the model training at GCP Vertex AI infrastructure. I might open source the steps to do that later, but now the code is messy. Pixray does have an accessible web UI at https://replicate.com/pixray/text2image if you are interested to try it out. You can't control all parameters there though. The most recent famous AI art generator is https://openai.com/dall-e-2/, which has a wait list.

This is a just-for-fun post. Over on the discord, Mylie and I have been getting heavily into GAN/CLIP style AI image generation, applied to pixel art with the pixeldrawer module of pixray. It has proven to be extremely fun to input Ultica art as a style guide and then use text prompts to generate new images. We've developed a new method for making seamless tiling pixel art, mixing the AI stuff with our own! Check these out:

This is also hardware-hungry I'm afraid, this uses Pixray here: https://github.com/pixray/pixray

Thanks! I'm using Pixray: https://github.com/pixray/pixray

Can you expand? This looks like something generated by pixray (https://github.com/pixray/pixray), is that what you tweaked? Looks really cool.

They are! And you will be able to generate an unlimited number of them in my game. It is based off this implementation. Hope that helps :)