style2paints
@lvminzhang
Followers
9K
Following
543
Media
135
Statuses
840
Interactivity, creativity and possibility!
Joined September 2017
RT @LTXStudio: Today we’re introducing our latest open source video model—and it’s a big one. This release sets a new bar for speed, quali….
0
204
0
VBench now seems to endorse that Open-Sora outperforms OpenAI's Sora.
6
6
97
4o-image passed this injection test. CLIPVision-like MLLM (like Janus-Pro-7B) cannot pass this test. But I am not absolutely sure.
1
1
19
It seems that you can do way more advanced "SQL injections"
0
4
24
Update2: it seems that the success rate is around 80%. Some images may fail.
0
2
7
Update 1: it seems that this method only works when the image is large enough, like width > 1024 pixels.
0
0
2
(To the best of my knowledge, this method seems to be proposed in this tweet. So, cite this tweet if necessary. Also, if one know any earlier sources that proposed the same, feel free to let me know.).
0
0
4
Probably the only correct method for artists to protect their artworks from being processed by the latest multi-modal LLMs like GPT-4o is to add this mark to your work:
9
21
65
After some investigation, the peak seems to be caused by normal user activity with different hf accounts and different user patterns.
IC-Light V2 HF demo is currently in a high workload. Not sure if this is increased need in illumination, or if someone is scraping training data, or if someone is wrapping the api to fake services … we will evaluate and take some action to improve experiences for normal users
5
0
25
IC-Light V2 HF demo is currently in a high workload. Not sure if this is increased need in illumination, or if someone is scraping training data, or if someone is wrapping the api to fake services … we will evaluate and take some action to improve experiences for normal users
7
4
55
Cuz people asked: the results in the GIFs are just simple direct outputs from iclight-v2 hf demo. Do not need any advanced settings or complex things.
4
12
127
To reproduce: prompts: “Cinematic lighting of a man. Light and shadow. Hard shadow from window.”; “Cinematic lighting of a woman. Sunshine from tree shade. Light and shadow” Use standard IC-Light V2 (not the -vary one) Images:
github.com
misc images
2
1
17
Alternative IC-Light V2 Model(s) for people who want stronger illumination variations and modifications.
github.com
Alternative IC-Light V2 Model(s) for people who want stronger illumination variations and modifications. Demo here: https://huggingface.co/spaces/lllyasviel/iclight-v2-vary Examples Input image: Ou...
11
46
215
If you ever, somewhat, think about having seen some reference transfer method discussions a long time ago by stitching images or inpainting, but just could not remember where you saw it exactly --- maybe here is that commit on 2023 May 13:
3
2
46
RT @Gradio: MagicSketch 🤩 . Interactive image editing Gradio app. 🤯 An MLLM even infers editing intent in real-time and generates a prompt….
0
40
0
A new virus is spreading on GitHub, disguised as "technical responses" in GitHub Issues. Once downloaded, it takes control of your computer and uses your GitHub account to respond to repos you've starred or bookmarked with virus. Many users and repos have already been hacked.
2
19
77
Congratulation to leejet: now native cpp cuda GGUF computation is coming to Flux
github.com
Although the architecture is similar to sd3, flux actually has a lot of additional things to implement, so adding flux support took me a bit longer. After merging this pr, I will take some time to ...
0
20
99