Kaizen
@kaizenteki.bsky.social
🏳️🌈 🫂🏳️⚧️
Tech / science / books / LGBT / music stuff I dig.
Tech / science / books / LGBT / music stuff I dig.
Reposted by Kaizen
Today is World AIDS day ❤️
World AIDS day has been marked every year since 1988, to:
❤️ Fight the stigma of HIV/AIDS
❤️ Show solidarity with those living with HIV/AIDS & honour those we have lost to AIDS-related illnesses
❤️ Raise awareness of the testing, treatments & interventions available
World AIDS day has been marked every year since 1988, to:
❤️ Fight the stigma of HIV/AIDS
❤️ Show solidarity with those living with HIV/AIDS & honour those we have lost to AIDS-related illnesses
❤️ Raise awareness of the testing, treatments & interventions available
December 1, 2025 at 1:40 PM
Today is World AIDS day ❤️
World AIDS day has been marked every year since 1988, to:
❤️ Fight the stigma of HIV/AIDS
❤️ Show solidarity with those living with HIV/AIDS & honour those we have lost to AIDS-related illnesses
❤️ Raise awareness of the testing, treatments & interventions available
World AIDS day has been marked every year since 1988, to:
❤️ Fight the stigma of HIV/AIDS
❤️ Show solidarity with those living with HIV/AIDS & honour those we have lost to AIDS-related illnesses
❤️ Raise awareness of the testing, treatments & interventions available
Konnakkol & Quads Transcription
YouTube video by percuss.io
www.youtube.com
November 30, 2025 at 4:45 PM
"My name is Alexander, King of Macedonia. I will grant you any request."
"Move over, you're blocking my sun."
"Move over, you're blocking my sun."
November 28, 2025 at 7:47 PM
"My name is Alexander, King of Macedonia. I will grant you any request."
"Move over, you're blocking my sun."
"Move over, you're blocking my sun."
Approximately 7% of American men think they can beat a grizzly bear in an unarmed fight, according to a 2021 YouGov survey.
November 27, 2025 at 8:34 PM
Approximately 7% of American men think they can beat a grizzly bear in an unarmed fight, according to a 2021 YouGov survey.
2025 internet in a nutshell - trust me bro just one more merger, if it ain't a monopoly then it's fine bro
November 18, 2025 at 2:21 PM
2025 internet in a nutshell - trust me bro just one more merger, if it ain't a monopoly then it's fine bro
World Labs has released Marble, a commercial world model that turns text, images, videos, and rough 3D layouts into persistent, editable 3D environments. The company is led by Fei-Fei Li, the Stanford professor known for creating ImageNet, the dataset that taught computers to see.
Fei-Fei Li's World Labs speeds up the world model race with Marble, its first commercial product | TechCrunch
Marble is different from competitors like Odyssey, Decart, and Google's Genie because it creates persistent, downloadable 3D environments rather than generating worlds on-the-fly as you explore.
techcrunch.com
November 15, 2025 at 6:13 AM
World Labs has released Marble, a commercial world model that turns text, images, videos, and rough 3D layouts into persistent, editable 3D environments. The company is led by Fei-Fei Li, the Stanford professor known for creating ImageNet, the dataset that taught computers to see.
Gpt-4: estimated to need $100 mln for training
Some later antropic model - took like 20-30 millions
Deepseek v3 - 8 millions or so
Some later antropic model - took like 20-30 millions
Deepseek v3 - 8 millions or so
November 8, 2025 at 2:11 PM
Gpt-4: estimated to need $100 mln for training
Some later antropic model - took like 20-30 millions
Deepseek v3 - 8 millions or so
Some later antropic model - took like 20-30 millions
Deepseek v3 - 8 millions or so
Spatial Intelligence: Photogrammetry, neural radiance fields (NeRFs), and 3D Gaussian splatting for digitizing reality.
Visual Intelligence: AI-powered pose estimation, segmentation, depth inference, and relighting.
... and more
Visual Intelligence: AI-powered pose estimation, segmentation, depth inference, and relighting.
... and more
The AI & 3D Tools No One Is Talking About (Harvard Lecture)
YouTube video by Bilawal Sidhu
www.youtube.com
October 18, 2025 at 9:59 PM
Spatial Intelligence: Photogrammetry, neural radiance fields (NeRFs), and 3D Gaussian splatting for digitizing reality.
Visual Intelligence: AI-powered pose estimation, segmentation, depth inference, and relighting.
... and more
Visual Intelligence: AI-powered pose estimation, segmentation, depth inference, and relighting.
... and more
So “prompting” at a high level is equal parts art direction, data engineering, and patience. It’s not typing fancy words into a box it’s designing and controlling a generative system down to the smallest variable.[7/7]
October 18, 2025 at 9:04 PM
So “prompting” at a high level is equal parts art direction, data engineering, and patience. It’s not typing fancy words into a box it’s designing and controlling a generative system down to the smallest variable.[7/7]
Doing it ethically makes things harder. Most artists refuse dataset licensing even when offered payment, so finding or creating legitimate material often becomes its own long, frustrating research project.[6/7]
October 18, 2025 at 9:03 PM
Doing it ethically makes things harder. Most artists refuse dataset licensing even when offered payment, so finding or creating legitimate material often becomes its own long, frustrating research project.[6/7]
That’s just for one or two characters. If the right LoRA doesn’t exist, you build one yourself: collect and clean a dataset, tag it properly, train the model, adjust learning rates, and hope it doesn’t overfit or collapse.[5/7]
October 18, 2025 at 9:01 PM
That’s just for one or two characters. If the right LoRA doesn’t exist, you build one yourself: collect and clean a dataset, tag it properly, train the model, adjust learning rates, and hope it doesn’t overfit or collapse.[5/7]
Once it renders, you’re nowhere near done. You inpaint bad spots, fix hands, correct lighting, adjust anatomy, balance colors, and sometimes re-render specific regions just to clean up the smallest inconsistencies.[4/7]
October 18, 2025 at 9:00 PM
Once it renders, you’re nowhere near done. You inpaint bad spots, fix hands, correct lighting, adjust anatomy, balance colors, and sometimes re-render specific regions just to clean up the smallest inconsistencies.[4/7]
Then comes structure control. ControlNet or T2I adapters for pose, depth, and composition. You feed it skeletons, edge maps, or references to make sure your characters don’t drift off or grow an extra arm mid-render.[3/7]
October 18, 2025 at 9:00 PM
Then comes structure control. ControlNet or T2I adapters for pose, depth, and composition. You feed it skeletons, edge maps, or references to make sure your characters don’t drift off or grow an extra arm mid-render.[3/7]
You write a base prompt and a negative prompt, select LoRAs or LyCORIS that match your style, add tags, balance their weights, fine-tune descriptions, and test different combinations until the model starts doing what you actually want. [2/7]
October 18, 2025 at 9:00 PM
You write a base prompt and a negative prompt, select LoRAs or LyCORIS that match your style, add tags, balance their weights, fine-tune descriptions, and test different combinations until the model starts doing what you actually want. [2/7]
AI “prompting” is way more than just typing words. If you want consistent, high-quality images, it’s a full creative pipeline involving composition control, model tuning, and constant iteration not a magic spell or a lucky roll that most believe it is.[1/7]
October 18, 2025 at 8:59 PM
AI “prompting” is way more than just typing words. If you want consistent, high-quality images, it’s a full creative pipeline involving composition control, model tuning, and constant iteration not a magic spell or a lucky roll that most believe it is.[1/7]
Reposted by Kaizen
For those who need to hear it today 🏳️🌈
January 20, 2025 at 7:43 AM
For those who need to hear it today 🏳️🌈
ちょっとリラックスしよう。
www.youtube.com/watch?v=1z56...
www.youtube.com/watch?v=1z56...
AGAiN - FlashGet kg_original - (.sunflower...)
YouTube video by Chiptune Dude
www.youtube.com
January 21, 2025 at 11:11 PM
ちょっとリラックスしよう。
www.youtube.com/watch?v=1z56...
www.youtube.com/watch?v=1z56...
2D CHARACTER IN A 3D BACKGROUND
2Dキャラクターが3D背景の中に
SPA Studios
2D CHARACTER IN A 3D BACKGROUND
Alfredo Cassano animates a 2d character in a 3d background. We hope you find this useful! Please leave any questions you may have in the comment section belo...
youtu.be
February 21, 2024 at 7:17 PM
2D CHARACTER IN A 3D BACKGROUND
2Dキャラクターが3D背景の中に
SPA Studios