Advertising on the Telegram channel «SmartSoftHub»

🪄 Image generation with Torch.FX Stable Diffusion v3 and OpenVINO 🛠 Stable Diffusion V3 is next generation of latent diffusion image Stable Diffusion models family that outperforms state-of-the-art text-to-image generation systems in typography and prompt adherence, based on human preference evaluations. 🔰In comparison with previous versions, it based on Multimodal Diffusion Transformer (MMDiT) text-to-image model that features greatly improved performance in image quality, typography, complex prompt understanding, and resource-efficiency. 🔗Links: https://github.com/openvinotoolkit/openvino_notebooks/tree/latest/notebooks/stable-diffusion-v3-torch-fx 🏷Tags: #stablediffusion #texttoimage

💻 Scrap 🛠 Scraperr is a self-hosted application designed to accurately extract data from websites using XPath selectors. 🔰 It provides a convenient interface for managing scraping tasks, viewing and exporting data. 🔰 Key features include XPath-based extraction, queue management, scraping all pages of a single domain, adding custom headers, automatic media downloading, and visualizing results in tables. 🔰 The application is intended only for sites where scraping is allowed, and the developer is not responsible for possible abuse. 🔗Links: https://github.com/jaypyles/Scraperr 🏷Tags: #scraping #website

🤩 OpenAI presented the Codex-1 model 🛠 Codex-1 is a cloud-based software development agent that can work on many tasks in parallel. 🔰Codex-1 was built on the o3 model, but generates cleaner code and handles a variety of tasks , including writing features and fixing bugs . 🤖 🔰Codex will iteratively test its code until passing results are achieved. 🔰The tool integrates with GitHub and runs in an isolated virtual cloud , providing access to user repositories . 🔰Codex is currently only available to paid ChatGPT Pro , Enterprise and Team subscribers, but will be expanded to other users (ChatGPT Plus and Edu) in the future . 💰 🔰Codex is intended as a "virtual companion" that helps engineers reduce the time spent on routine tasks . 🔰OpenAI is also actively working to improve the security of the Codex by restricting its access to external sources to minimize the risks of abuse . 🔒 🔗Links: https://openai.com/index/introducing-codex/ 🏷Tags: #openai #codegenerator

🎬 Higgsfield Video Generator Adds Opportunity to Become Part of Cult Movies 🛠The platform's creators have introduced new presets based on iconic movie moments and popular memes. 🔰New features include Iconic Scenes , which lets users insert themselves into iconic movie scenes , and a GPT-4o-powered image generator that transforms the style of images🖼 🔰Users can create up to two videos of up to five seconds for free , with paid subscriptions starting at $9 per month . 💰 🔗Links: https://higgsfield.ai/ 🏷Tags: #ai #videocreate

😎 ChatGPT has a new model GPT-4.1, optimized specifically for working with code. 🔰The model is far ahead of GPT-4o in programming and follows prompts much better. In everyday tasks, it easily replaces the “heavier” o3 and o4-mini. 🔰Previously, it was only available to developers via API. 🔰GPT-4.1 is only available to users with paid ChatGPT Plus, Pro, and Teams subscriptions .

🖥 Plexe 🛠 Plexe is a system that simplifies the creation of machine learning models by allowing users to describe their intent in natural language. 🔰 The system automatically generates functional models and is available as a Python library and cloud service. 🔰 Custom models are defined through a description including intent and input-output patterns, and can be created by a single team with support for distributed training using Ray. 🔰 Additionally, Plexe offers synthetic data generation and automatic schema inference features, supporting various LLM providers such as OpenAI and Anthropic. 🔗Links: https://github.com/plexe-ai/plexe 🏷Tags: #python #machinelearning

💥 Meta Announces Meta Perception Language Model (PLM) 🛠 PLM is an open and reproducible vision language model that solves complex visual problems by integrating image, video and text processing. 🔰 PLM was trained on the largest existing dataset with 2.5 million human-annotated videos. 🔰 The model is capable of creating detailed text descriptions for images and videos, taking into account not only objects, but also their actions, context, and relationships between them.🤖 🔰 PLM is distributed with 1, 3 and 8 billion parameters and is available to everyone on Github and HuggingFace. 🔗Links: https://github.com/facebookresearch/perception_models https://huggingface.co/collections/facebook/perception-lm-67f9783f171948c383ee7498 🏷Tags: #meta #ai

😎 The world's first AI browser Fellou with a built-in smart agent has been released 🛠 Fellow — a browser with a built-in AI assistant. The developers call it not just a browser, but a full-fledged agent that understands the user's tasks and performs them without unnecessary fuss ⚙️ Some features of Fellou: 🔹 Control in the “say what you want” format. For example, you don’t need to look for a link to YouTube – just ask to open it; 🔹 Deep analytics tool. AI can read multiple tabs at once, compare the content and compile a clear report from it; 🔹 Memory: the browser remembers style, behavior and typical scenarios. Over time, Fellou begins to predict what the user wants to do: for example, it will offer to open the necessary tabs by the beginning of the working day or create a draft letter using a template. 🔗Links: https://fellou.ai/ 🏷Tags: #ai #browser

📈Rybbit 🛠 Rybbit is a modern, privacy-focused web analytics tool that acts as an alternative to Google Analytics. 🔰 The platform offers both a cloud version for a quick start and a self-hosting option for users seeking full control over their data. 🔰 Supported metrics include sessions and unique users, all without the use of cookies. 🔰 Compliant with GDPR and CCPA requirements, Rybbit provides goal customization, dashboards, and advanced data filtering, allowing users to get visualizations and data in real-time. 🔗Links: https://github.com/rybbit-io/rybbit 🏷Tags: #website #traffic #analytics

📹 Lightricks Introduces LTXV-13B Video Generation Model for Home PCs 🛠 With 13 billion parameters, this model creates content 30 times faster than similar solutions on the market. 🔰Featuring innovative “multi-scale rendering” technology , the LTXV-13B runs smoothly on consumer devices including laptops with RTX 3090 , 4090, and 5090 graphics cards. 🔰The source code is available on Hugging Face and GitHub platforms , providing free access to startups. 🔗Links: https://github.com/Lightricks/LTX-Video https://huggingface.co/Lightricks/LTX-Video 🌐Site: https://ltxv.video/ 🏷Tags: #ai #videogeneration #huggingface