Генерация видео: Обзор интересных подходов | Text-2-video | Part 1
План следующий: Методы адаптации T2I в T2V : AnimateDiff, Text2Video Zero Обзор классических подходов : Stable Video Diffusion, CogVideo Новые модельки 2025 : Wan2.1, HunyuanVideo, FramePack Это первая часть из списка статей, тут будет только про T2I в T2V
https://habr.com/ru/articles/910326/
#computer_vision #нейросети #ml #video_generation #собеседования #собеседования_задачи #ai #stable_diffusion #comfyui #animatediff
New substack post taking a look at some of the current video diffusion models available in ComfyUI and online
https://diffusiondoodles.substack.com/p/big-steps-forward-in-video-generation
Trying out Kijai's Flux Trainer nodes in #comfyui, overall a great experience so far, I've never been a huge Kohya fan (mostly due to the UI), so this works a lot better for me.
AI oder nicht AI? Das ist hier die Frage. Nur diesmal auf Bossfight-Level: Power Rangers Figur oder generativer AI-Schrott?
Der Insight kam von meiner Frau: „Diese Power Rangers Monster sehen aus wie miese KI Bilder." Zack fertig: ein derbe weirdes AI Quiz.
👉 https://butschcoolidge.github.io/aipowerrangersquiz/
Und – schafft ihr 20?
#flux #aiquiz #comfyui #aiornotai #powerRangerOrPromptAbfall
🎬 We've released a fresh video in the "GenAI's Lamp" series, spotlighting ComfyUI's Native API Nodes with OpenAI GPT.
Perfect for those eager to explore AI-driven image generation and style shifts.
From precise prompts to creative outputs, this tutorial is packed with insights to elevate your projects. Dive in and see GenAI in action! 🔮
Check it out now! https://youtu.be/NxQA-DN4FAA
🎶 On yesterday rainy Sunday, I explored ACE-Step, a new open music model. It's easily installed with #ComfyUI to create unique melodies.
While not yet matching proprietary tools like Suno, ACE-Step is revolutionary—akin to Stable Diffusion for images. It supports control features like regenerating variations, reworking parts, and editing lyrics. An exciting start to collaborative musical creativity!
Can't wait to see community advancements.
#GenAI #ACEStep #MusicGeneration #TechAtWorldline
CLIP или SigLIP. База по Computer vision собеседованиям. Middle/Senior
Вопросы о CLIP-моделях встречаются почти на каждом техническом собеседовании. Неважно, занимаетесь ли вы видеоаналитикой, создаёте генеративные модели или работаете над поиском по изображениям — CLIP и его потомки ( BLIP , SigLIP ) стали стандартом де-факто в задачах связи визуальных и текстовых данных. Почему? Потому что они позволяют решать задачи, которые ранее требовали значительных усилий
https://habr.com/ru/articles/908168/
#clip #SigLIP #компьютерное_зрение #computervision #ml #машинное+обучение #собеседование_вопросы #собеседование_в_it #comfyui
🔥 #ComfyUI gets a "Branding New Look" & MASSIVE API power-up! ✨
Now natively integrate 11+ top visual AI models:
→ Incl. Flux, Hailuo, Kling, Luma, Veo2, GPT image-1, Pika & more!
→ One platform: 11 models, 65 nodes at once.
→ Chain these APIs with open models/workflows.
→ No hardware limitations!
Ready to supercharge your ComfyUI workflows?👇🏻
If you’re running #ComfyUI on an Apple Silicon Mac (M1/M2), you might have encountered the frustrating error message: “Trying to convert Float8_e4m3fn to the MPS backend but it does not have support for that dtype“. This comprehensive guide will walk you through understanding and solving this error, ensuring smooth operation of your AI workflows on Apple Silicon.
Solutions:
- fp8_unet to false in node (my case)
- Force fp16 (common case)
HiDream-I1 isn't perfect and occasionally fails in similar ways to Flux but its prompt adherence is strong. The images just look different to Flux but hard to say what is different!
This image was created with no LoRAs or refined models, just straight up HiDream-I1 (Q8 GGUF) with multi-stage sampling and Ultimate SD Upscale.
Experimenting with the Q8 quantised version of Hi-Dream. Performance/quality trade-off is good.
This week's post on the Diffusion Doodles Substack (https://diffusiondoodles.substack.com/p/how-to-train-a-lora) looks at creating your own LoRAs for consistent styles and characters.
#GenAI #diffusionmodels #imagegeneration #comfyUI #flux #loras #characterdesign
Больше 5090 — больше проблем? Тестируем связку из двух GPU NVIDIA
В предыдущей статье мы рассказывали, как тестировали сервер с одной RTX 5090. Теперь же мы решили установить в сервер две видеокарты RTX 5090. И это также вызвало у нас определённые проблемы, но результат того стоил.
https://habr.com/ru/companies/hostkey/articles/903194/
#nvidia #5090 #ollama #comfyui #openwebui #видеокарты #серверное_железо
📢 Upset about losing CVE data? We dropped these new 🔥 critical #ComfyUI vulnerabilities, including this RCE, to cheer you up 😉. These are published as part of our coordinated disclosure process.
https://www.doyensec.com/resources/Doyensec_Advisory_ComfyUI_Manager_RCE_via_Custom_Node_Install.pdf
Narrated Noir
🎥 I'm transitioning from music videos to storytelling with my latest project: Narrated Noir!
AI video tech is evolving fast. Home creators work unrestricted, pushing boundaries the big players can’t without bigger costs.
Follow my progress and see how it’s done.
📍 Website: markdkberry.com
🎶 Music: markdkberry.bandcamp.com
📸 IG: @markdkberry
📘 FB: facebook.com/markdkberry.productions
✖️ X: @markdkberry
an Inhabitant of Carcosa https://youtu.be/53pmJRcfV0M?si=731ThFyh-63rIFch via @youtube #4K, #60fps, #UHD, #horror, #animation, #surreal, #Lovecraft, #fantasy, #supernatural, #occult, #comfyui, #LatentSync, #HunyuanVideo, #mystery, #Carcosa, #Hastur, #cosmichorror, #Cthulhu
