Projects

Overview: ML Guide is a comprehensive repository designed for machine learning enthusiasts and professionals alike, offering a wide range of resources, tutorials, projects, and datasets to enhance learning and skill development. Key features: - A variety of projects and code samples for practice and real-world experience. - Open-source, allowing anyone to contribute and modify the content. Tech stack: Jupyter Notebook Why it matters: ML Guide serves as a valuable resource for anyone interested in machine learning, providing a structured and collaborative environment for learning and skill enhancement.

Overview: This open-source marketing skills repository offers free tools for founders to improve their marketing strategies, including social search audit, growth strategy, keyword research, and competitor analysis. Key features: - Social Search Audit: Evaluates TikTok/Instagram searchability for growth. - Growth Strategy: Generates a comprehensive marketing strategy. - Keyword Research: Finds high-intent keywords for your niche. - Competitor Audit: Identifies gaps and positioning angles for competitors. Tech stack: Skills are installed via the CLI with `npx skills add`. Why it matters: These tools help founders optimize their marketing efforts, targeting high-intent keywords and analyzing competitors, leading to more effective and efficient strategies. Repository: ScaleBrick/founder-marketing-skills GitHub stars: 88 GitHub description: Open-source marketing skills for founders.

Overview: Omni-Rewriter is an open agentic prompt-expansion harness designed for image and video generation. It bridges polished demos, public APIs, and deployable workflows. Key features: - Bounded AI-agent workflow - Public LTX-2 prompting guide PE - Recommended Writer: Qwen3.6-35B-A3B - H3 video PE workflow with validated timeline, camera, dialogue, and bounded-repair rules Tech stack: Python Why it matters: Omni-Rewriter provides a seamless and efficient way to generate high-quality images and videos by leveraging AI-powered prompt expansion, making it accessible and practical for developers and creators alike. Repository: WayneJin0918/Omni-Rewriter Primary language: Python GitHub stars: 89 GitHub description: Open agentic prompt-expansion harness for image and video generation, bridging polished demos, public APIs, and deployable workflows. README excerpt:

### Overview: Creating multi-view wire art (MVWA) is a challenging task for skilled artists. *DreamWire* is an AI system that enables anyone to easily create MVWA by expressing their vision through text prompts or scribbles. This is the first AI solution for generating multi-view wire art. ### Key Features: - **Text Prompt Input**: Users can express their vision through text prompts or scribbles. - **Generative AI**: Utilizes generative AI to create diverse interpretations of MVWA from different viewpoints. - **Hybrid Mode**: Supports generating MVWA in a hybrid mode where text prompts control the sketching process. ### Tech Stack: - **Python**: The primary programming language used for developing *DreamWire*. ### Why It Matters: *DreamWire* simplifies the process of creating MVWA, making it accessible to artists of all skill levels. By leveraging generative AI, it provides a new way to interpret and visualize complex ideas, potentially opening up new creative possibilities in wire art.

Overview: LiDPM is an official code repository for the paper "LiDPM: Rethinking Point Diffusion for Lidar Scene Completion," presented at IEEE IV 2025. The project includes training scripts, inference tools, and a SemanticKITTI dataset for evaluation. Key features: - Training scripts for diffusion models - Inference tools for completing lidar scans - SemanticKITTI dataset for validation and evaluation Tech stack: Python Why it matters: LiDPM provides a comprehensive solution for scene completion using diffusion models, offering a practical approach to lidar scene understanding and reconstruction. Repository: astra-vision/LiDPM Primary language: Python GitHub stars: 89

### Overview: TIME (Text-to-Image Model Editing) is an innovative method designed to edit implicit assumptions in text-to-image diffusion models. This project introduces a new approach to modifying text-to-image diffusion models, enhancing their ability to generate realistic images with specific characteristics. ### Key Features: - **Text-to-Image Model Editing**: TIME enables users to modify the implicit assumptions in text-to-image diffusion models, making them more versatile and adaptable. - **Gradio Demo**: A browser-based demo is available, allowing users to experiment with TIME directly from their browser. - **GitHub Stars**: The repository has been starred by 89 developers, indicating its growing interest and utility in the community. ### Tech Stack: - Python - PyTorch 1.13.1 - CUDA 11.6.2 - Numpy 1.23.4 - Diffusers 0.9.0 ### Why It Matters: TIME represents a significant advancement in the field of text-to-image diffusion models, providing a powerful tool for enhancing the capabilities of these models. Its ability to edit implicit assumptions makes it particularly useful for applications requiring precise control over generated images, such as in the creation of realistic and contextually accurate images for various industries and applications.

Overview: This repository contains the source code for the paper "Diffusion-Based Generative Prior for Low-Complexity MIMO Channel Estimation," which proposes a novel channel estimator based on diffusion models for low-complexity MIMO channel estimation. Key features: - Lightweight CNN with positional embedding for learning channel distribution - DMSE scheduler for efficient estimation without stochastic resampling or truncation of reverse diffusion steps - Pre-trained models for evaluating and training estimators Tech stack: Python Why it matters: This project offers a practical solution for low-complexity MIMO channel estimation, providing a more efficient and accurate alternative to state-of-the-art estimators. GitHub stars: 89

Overview: Callytics is an advanced call analytics solution that uses speech recognition and large language models to analyze customer service and call center conversations. It provides insights like sentiment analysis, topic detection, conflict detection, and profanity word detection. Key features: - Sentiment analysis - Topic detection - Conflict detection - Profanity word detection Tech stack: Python Why it matters: Callytics helps businesses optimize customer interactions, identify areas for improvement, and enhance overall service quality by processing both audio and text from phone conversations. Repository: bunyaminergen/Callytics GitHub stars: 90 GitHub description: Callytics is an advanced call analytics solution that leverages speech recognition and large language models (LLMs) technologies to analyze phone conversations from customer service and call centers. By processing both the audio and text of each call, it provides insights such as sentiment analysis, topic detection, conflict detection, profanity word detection and summary. These cutting-edge techniques help businesses optimize customer interactions, identify areas for improvement, and enhance overall service quality. License: Not specified in the provided excerpt.

Overview: Grok Imagine 2.0 API is a powerful tool for generating high-quality images using OpenAI's models. It offers flexible pricing and region-specific options, making it accessible to developers and creators alike. Key features: - Upload an image for $0.015 - Generate default image for $0.015 - Edit region-specific image for $0.015 Tech stack: Shell Why it matters: This API provides a straightforward and efficient way to generate images, making it ideal for developers and creators looking to integrate image generation into their projects without the complexity of setting up their own models. Repository: bx0idu7sqebwpgu8/grokimagine2.0-grok-imagine-2.0-prompts Primary language: Shell GitHub stars: 89

### Overview: The GPT Image 2.5 API offers high-resolution images generated by the GPT model, with prices starting at $0.0085 per unit for 1K resolution and increasing to $0.014 per unit for 2K resolution. This API is compatible with OpenAI's API and supports both flat per-unit billing and API key-based pricing. ### Key Features: - **1K Resolution (Flare)**: $0.0085 per unit - **1K Resolution (Sunburst)**: $0.0085 per unit - **2K Resolution (Flare)**: $0.014 per unit - **2K Resolution (Sunburst)**: $0.014 per unit ### Tech Stack: - Shell ### Why It Matters: This API provides high-quality images generated by the GPT model, making it suitable for various applications such as marketing, design, and more. The pricing is flexible, offering both flat per-unit billing and API key-based pricing, allowing for cost-effective usage.

Overview: This extension allows developers to integrate Grok functionality with Twikit, enhancing the ability to analyze and respond to text-based conversations. Key features: - Streaming and non-streaming conversation handling - Support for generating responses and attachments - User-friendly installation and quick example Tech stack: Python Why it matters: This extension streamlines the process of integrating Grok with Twikit, making it easier for developers to analyze and respond to text-based conversations in their applications.

Overview: This project is an official implementation of the CVPR 2025 paper "Progressive Autoregressive Video Diffusion Models". It includes the training and inference code for the PA-VDM model. Key features: - Progressive Autoregressive Video Diffusion Models - Open-Sora v1.2 training and inference code - Testing data of 40 real videos and text prompts - New training/inference configurations for "Chunked Frames", "Overlapped Conditioning", and "Variable Length" Tech stack: Python Why it matters: This project provides a practical implementation of the PA-VDM model, which is designed for progressive autoregressive video diffusion. It includes comprehensive documentation and testing data, making it a valuable resource for researchers and developers interested in video diffusion models.

### Overview: B.E.N.J.I. is a digital assistant designed for civilian use, capable of executing commands through speech recognition. It was developed as part of various Open-Source Programs and is now available on Windows, Linux, and Mac OS. ### Key features: - Speech recognition for command execution - Works on Windows, Linux, and Mac OS - Developed as part of Kharagpur Winter of Code 2017, GirlScript Summer of Code 2018, NJACK Winter of Code 2018, and Kharagpur Winter of Code 2018 - Featured at the Open-Source Summit held at IIT Kharagpur in January 2018 ### Tech stack: Python ### Why it matters: B.E.N.J.I. is a tool designed to assist with civilian operations, ensuring its use is peaceful and controlled. It is a result of various Open-Source Programs and is now available on multiple operating systems, making it accessible to a wider audience. Repository: dhruvapte26/B.E.N.J.I. GitHub stars: 89

Overview: OpenMelon is a content-creation agent that simplifies the process of generating and managing content, making it accessible through a terminal session. Key features: - Project management: Keeps projects and their characters, references, and generated artifacts on disk. - Skill integration: Anchored to the same character portraits across runs, using a skillplus package and LLM. - Direct prompt vs OpenMelon: Differentiates between sending user intent directly to the image model and expanding it through a skillplus package and LLM. - Installation: Available via npm package or building from source. - Configuration: Walks through trust, API key, LLM model, image model, and project initialization. - TUI commands: Provides access to a command palette for skill switching, model switching, and session management. Tech stack: Go Why it matters: OpenMelon streamlines the content creation process, making it easier and more efficient to generate and manage content through a terminal interface, leveraging skills and LLMs for enhanced functionality. Repository: eight-acres-lab/openmelon GitHub stars: 89 GitHub description: A content-creation agent runtime for reproducible multimodal production — projects, workflows, skills, provenance.

### Overview: Pywhisper is a Python library that extends the functionality of the Whisper model by adding several key features and improvements. It simplifies the installation process, eliminates the need for external dependencies like FFmpeg, and supports continuous integration and package testing via GitHub Actions. ### Key Features: - **Easy Installation**: Pywhisper can be installed from PyPI, making it straightforward to get started. - **No FFmpeg Dependency**: It requires only Python and pip, eliminating the need for FFmpeg. - **Continuous Integration**: GitHub Actions ensure that the library is tested and packaged regularly. - **Continuous Testing**: Package testing is performed to ensure the library's quality and stability. ### Tech Stack: The library is built using Python, with a focus on the Whisper model from OpenAI. It leverages Rust for tokenizers, which is essential for the Whisper model's performance. ### Why It Matters: Pywhisper simplifies the process of using the Whisper model by providing a user-friendly interface and ensuring that the library is robust and reliable. Its continuous integration and testing practices make it a valuable tool for developers looking to integrate Whisper into their projects without worrying about installation or compatibility issues. ### Repository: The repository is maintained by fcakyon and is currently at GitHub stars 89. It is a valuable resource for anyone working with speech recognition models, offering a convenient and reliable way to use the Whisper model.

### Overview: Pluto is a powerful platform that allows developers to run AI agents within Salesforce, utilizing standard licenses. It includes a robust engine, model adapters, and a comprehensive console for managing agents and operations. ### Key Features: - **Pluto Core**: The engine that runs, tools, approvals, recovery, and data protection. - **Pluto Model Adapters**: Ready-made OpenAI connections. - **Pluto Console**: Admin console for setting up agents, viewing runs, and making approvals. ### Tech Stack: Apex ### Why it Matters: Pluto simplifies the integration of AI agents into Salesforce, making it accessible without requiring expensive licenses. It handles model calls, tool approvals, and recovery, ensuring a seamless and secure experience for developers and users alike.

### Overview: EmbraceNet is a robust deep learning architecture designed for multimodal classification tasks. This innovative method integrates various modalities effectively, ensuring compatibility with any network structure and handling missing data seamlessly. The repository includes official PyTorch and TensorFlow implementations of the EmbraceNet model, detailed in a paper presented at the SHL recognition challenge 2019. ### Key features: - **Robust Multimodal Integration**: EmbraceNet seamlessly combines multiple modalities, enhancing model performance. - **Compatibility with Any Network**: The architecture is compatible with various network structures, ensuring flexibility. - **Handling Missing Data**: It seamlessly handles missing data, maintaining model accuracy. ### Tech stack: - **Python 3.7+** - **PyTorch 1.5+** - **TensorFlow 2.4+** ### Why it matters: - **Versatile Application**: EmbraceNet excels in multimodal classification tasks, offering a versatile solution for various applications. - **Performance Enhancement**: The architecture significantly improves model performance compared to traditional methods. - **Comprehensive Documentation**: Detailed comments in the code snippets provide clear guidance on how to implement and use EmbraceNet effectively. ### Repository: - **GitHub Stars**: 89 - **Primary Language**: Python - **Dependencies**: - **PyTorch-based**: Python 3.7+ | PyTorch 1.5+ - **TensorFlow-based (2.x)**: Python 3.8+ | TensorFlow 2.4+ - **TensorFlow-based

### Overview: The FLUX.2 Pro API (flux-2 / flux2) gateway guide offers flat per-unit billing for 1MP, 2MP, and 3MP, with a minimum top-up of $1. The API supports OpenAI-compatible APIMart gateway, providing $0.024 per 1MP, $0.036 per 2MP, and $0.048 per 3MP. The pricing is observed and can be accessed through the live pricing link. ### Key features: - **Pricing**: $0.024 per 1MP, $0.036 per 2MP, $0.048 per 3MP, with a minimum top-up of $1. - **OpenAI-Compatible Gateway**: Supports OpenAI-compatible APIMart gateway for secure and efficient API calls. - **Documentation**: Includes detailed documentation on how to call the API using `curl` and Python examples. - **Model ID**: Supports model ID for more precise control over the generated content. - **Gateway Settings**: Allows for setting gateway settings for optimal performance. ### Tech stack: Shell ### Why it matters: This API gateway is crucial for developers looking to integrate the FLUX.2 Pro model into their projects, offering flexible pricing and a reliable OpenAI-compatible gateway for secure API calls.

### Overview: NOVA is a customizable voice assistant built with Node.js, designed to be hosted on your server for personal use, offering offline capabilities to protect privacy. ### Key Features: - **Voice and Text Interaction**: Communicate with NOVA both orally and in writing. - **Customizable Skills**: Install and manage skills to enhance its functionality. - **Server and Client**: Hosted on a server for better performance, accessible via web interface. - **Performance Issues**: The project has been identified as having performance issues on Raspberry Pi, requiring a Chromium-based browser to work well. ### Tech Stack: Node.js ### Why it Matters: NOVA is a personal assistant that can be hosted on your server, providing offline capabilities to protect your privacy. The project is open-source and can be customized to fit your needs, making it a valuable tool for personal use.

### Overview: This repository contains a script designed to generate multi-modal speech separation data for the LRS3 dataset, focusing on the task of pure voice separation. The script is intended to provide a standardized approach for data generation, aiming to facilitate further research and benchmarking in the field of multi-modal speech separation. ### Key Features: - **Data Generation**: The script is capable of generating both training and testing data for the LRS3 dataset. - **Parameter Tuning**: Users can adjust various parameters to suit different purposes, ensuring the data is tailored to specific needs. - **Unified Standard**: The repository aims to establish a unified standard for generating multi-modal speech separation tasks, similar to the WSJ0 dataset for pure voice separation. ### Tech Stack: - MATLAB ### Why It Matters: This script provides a standardized approach for generating multi-modal speech separation data, making it easier to follow up on research in the field. The unified standard ensures that future studies can build upon the same data generation methods, facilitating reproducibility and benchmarking. ### Repository: - **GitHub Stars**: 88 - **Primary Language**: MATLAB - **GitHub Description**: Multi-modal speech separation task data generation script on LRS3 data set. - **README Excerpt**: The repository includes steps for generating training and testing data, using the LRS3 dataset for pure voice separation tasks. The script is designed to be easily adaptable to different parameters and benchmarks the performance of baseline models. ### Requirements: - **ffmpeg 4.2.1** - **sox 14.4.2