sahil_chaudhary

6 ms·

31.
by sahil_chaudhary 4y ago
The model was trained exclusively on stable diffusion generated images, so can be unpredictable with non-generated images, specially images with people in it
32.
by sahil_chaudhary 4y ago
The dataset used to train this model didn’t have any img2img data so that would explain it
33.
by sahil_chaudhary 4y ago
Agreed, results not good for this style of images. I have a model training on a much bigger dataset of image-prompt pairs which should perform better on this.
34.
by sahil_chaudhary 4y ago
The model underneath this is trained only on data from SD 1.4/5 so this would be expected. I have another model training which covers all models you mentioned which should perform well on those
35.
by sahil_chaudhary 4y ago
Interesting, it hasn’t been tested extensively for non-generated images.
36.
by sahil_chaudhary 4y ago
Not yet, looking into creating a write up on how it works and possibly open-sourcing
37.
by sahil_chaudhary 4y ago
Shoutout to https://banana.dev , couldn't have made this demo without their hosting
38.
by sahil_chaudhary 4y ago
It is actually based on a different approach, it uses an image-captioning model finetuned on image-prompt pairs
39.
by sahil_chaudhary 4y ago
Glad you like it
40.
by sahil_chaudhary 4y ago
Nice, that's the aim, not to get the actual prompt back but get a prompt which can generate same style of images
41.
by sahil_chaudhary 4y ago
Glad you enjoy it
42.
by sahil_chaudhary 4y ago
True, images generated through some UIs have prompts in meta data, aim here is to work on images people find online with no metadata. So it doesn't try to read the metadata but actually predict a similar prompt
43.
by sahil_chaudhary 4y ago
It is based on an image-captioning model, so a different approach then CLIP interrogator, though you are correct that aim is not to get the exact prompt back but actually get a prompt to generate similar styles of images
44.
by sahil_chaudhary 4y ago
It works by using an image-captioning model finetuned on SD prompts, so it may be outputting common known artists based on their occurrence in the training data
45.
by sahil_chaudhary 4y ago
It actually works on top of an image captioning model, SD takes in keywords as well like "artstation" and "octane render" which are not covered in standard captioning so that is why the difference between using an off-th
46.
Img2Prompt – Get prompts from stable diffusion generated images(img2prompt.io)
183 points by sahil_chaudhary 4y ago|75 comments
47.
Reverse engineer Stable Diffusion prompts(img2prompt.io)
1 points by sahil_chaudhary 4y ago|0 comments
48.
by sahil_chaudhary 4y ago
tested with png and jpg, doesn't work with webp yet
49.
by sahil_chaudhary 4y ago
Hey HN! Img2Prompt lets you generate prompts from stable diffusion generated images, to help you prompt engineer specific styles and images. Would love to hear your thoughts and happy to answer questions.
50.
Show HN: Img2Prompt – Tool to generate prompts for Stable Diffusion images(img2prompt.io)
3 points by sahil_chaudhary 4y ago|3 comments
51.
by sahil_chaudhary 4y ago
Hey all! Quick colab to try out Img2Prompt, get accurate prompts from stable diffusion generated images. Works with stable diffusion v1.4/5 images for now
52.
Show HN: Img2Prompt, Get prompts from Stable Diffusion generated images(colab.research.google.com)
15 points by sahil_chaudhary 4y ago|1 comments
53.
by sahil_chaudhary 4y ago
Adding support for sending images to chatGPT using the carrot image captioning model
54.
Multimodal ChatGPT bot – Sending images to ChatGPT(github.com)
2 points by sahil_chaudhary 4y ago|2 comments
55.
by sahil_chaudhary 4y ago
I think it’ll need little prompt engineering to figure out what works best but should be able to generate pretty good after that
56.
by sahil_chaudhary 4y ago
Yes, so people can freely try it out
57.
Make music with AI using Riffusion, without a GPU(colab.research.google.com)
3 points by sahil_chaudhary 4y ago|5 comments