#SDXL

Nicolas MOUARTsilentexception
2025-11-05

Dream designer, 2025.

Art deco AI generated image of a woman at her desk in front of a screen with a sort of web on it, but it is not a spider web, more like a dream catcher from Native American heritage. She seems lost in her thoughts or thinking about something.

NB: see Qwen3 VL 30B A3B Q4 alt text in alt text. Created in 🇫🇷 under 170W.

#alttext

CC BY-SA 4.0 // AI generated

A woman with dark, curled hair sits at a desk in a dimly lit room, gazing at an ornate, vintage-style television with a circular screen. She wears a sleeveless, draped dress and layered necklaces, her hands resting on the desk near a small, intricate globe. The scene is rich with art deco and steampunk elements, including a bookshelf filled with books, a patterned chair, and decorative framed artwork on the wall.  - Qwen3 VL 30B A3B Q4
2025-11-04

Xây dựng trạm làm việc AI nhỏ với 2× RTX 5090 cho SDXL, sinh video và suy luận LLM thường xuyên. Tìm kiếm giá trị tốt nhất cho hiệu suất #AI #SDXL #LLM #GPU #TrạmLàmViệc #Intel #NVIDIA #RTX5090 #SinhVideo #SuyLuậnLLM #TrựcTiếpTrênPrem

reddit.com/r/LocalLLaMA/commen

Ohm RaumzeitOhm
2025-10-31

WARNING about the malicious code called "Reasoning", which is added to LLM, GPT and AI from Moscow and interposed as a man-in-the-middle. github.com/ohm-raumzeit/itheer

Nicolas MOUARTsilentexception
2025-10-30

Social media streamers keep talking about speed for , it is wrong: who needs a system that produces subpart outputs that one needs to painstakingly verify by hand? 6GB is enough. Cloud ones are not accurate either: can't be trusted at all sizes.
Summary:
- -OSS-20B Q8 works @ 5 tok/s, not accurate, reasoning take lots of tok.
- VL 30B A3B Q4, same @ 5 toks/s.
- is fine, relatively fast, great for inspiration.
- , not enough, very slow.
NB: 32GB ram required.

A wattmeter showing power consumption. The Nvidia A2000 RTX 6GB is 70W, under 170W total depending on task.screenshot from ComfyUI showing the image of a pelican sitting on a branch, and some artic animal on an iceberg.

Little update on my #neuralnetwork #experiment: Something happened. What happened is hard to say so far. The motionnet model was only trained on 3 videos for 2000 steps. One of the videos was a butterfly on a flower. (You can see it attached).

So this is the first actual test I've run successfully on the full pipeline. The prompt: A butterfly on a flower.

You can see the first frame (generated by #SDXL weights) and the generated #video.

A butterfly with orange and black patterned wings is perched on a small purple flower, with green leaves and a blurred background.

Been messing around with a little #prototype #neuralnetwork. It behaves similarly to a #video #codec. It uses a frozen set of weights from #StableDiffusionXL and uses its latent space to carry forward to a set of networks that behave like the B-frame motion vectors used by codecs like #MPEG. These are smaller networks, letting me train it on my regular old #GPU while relying on the work of the big boys for generating I-frames via #SDXL.

At least that's the theory. Reality remains to be seen.

Terminal output showing a training log entry with the timestamp “2025-10-13 23:59:52,603,” indicating an INFO message for a training step. The log line includes “step=50,” “total=16669.1611,” “reconstruction=22760.1660,” and “temporal=23908.9961,” followed by “611.78s.” A progress bar below shows “66/2000” steps completed, with an estimated time of 11 hours, 16 minutes remaining.
2025-10-14

生成AIグラビアをグラビアカメラマンが作るとどうなる?第52回:オープン画像生成AIが怒涛の登場果たした2025年9月(西川和久)
techno-edge.net/article/2025/1
#technoedge #テクノロジー #ニュース #レビュー #ゲーム #ガジェット #生成AIグラビアをグラビアカメラマンが作るとどうなる #FLUX_1 #SDXL #LLM #LoRA

2025-09-07
2025-09-07

safe, 1girl, wicke (pokemon), pokemon, pokemon sm, playboy bunny, sitting, simple background, masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

2025-09-07
2025-09-06

safe, 1girl, lina inverse, slayers, plugsuit, standing, day, outdoors, masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

2025-09-06

safe, 1girl, milim nava, tensei shitara slime datta ken, rain jacket, standing, sky, 2000s (style), masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

2025-09-06

safe, 1girl, milim nava, tensei shitara slime datta ken, school uniform, standing, classroom, sketch, line art, masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

2025-09-06

safe, 1girl, byleth \(fire emblem\), fire emblem, plugsuit, lying, simple background, 1980s (style), masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

2025-09-06

safe, 1girl, , original, blue skin, see-through skin, slime girl, monster girl, school uniform, crossed legs, cafe, blurred background, masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

2025-09-06

safe, 1girl, , original, black hair, short hair, cat ears, :3, maid headdress, black dress, apron, black shoes, squatting,, masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

2025-09-06
2025-09-02

safe, 1girl, marie (splatoon), splatoon, plugsuit, standing, cowboy shot, simple background, sketch, line art, masterpiece, high score, great score, absurdres | #AIart #aiイラスト #AI #SDXL

Client Info

Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst