Projects

### Overview: This project provides a simple package for image editing using SAM, CLIP, and DIFFUSION, allowing users to modify objects in images with plain text instructions. ### Key features: - Image inpainting using SAM and CLIP - Modifying objects in images with text prompts - Easy installation via `pip` - Examples provided for both installation and without installation ### Tech stack: Python ### Why it matters: This project simplifies the process of image editing by abstracting away the complexities of using SAM, CLIP, and DIFFUSION, making it accessible to developers and artists alike.

### Overview: TRIBE v2 — Multimodal fMRI Brain Encoding Model Inference in Rust ### Key Features: - **End-to-End Media Pipeline**: Load raw media (video, audio, text) and predict brain responses in one command. - **GPU Inference via `--backend burn-gpu`**: 84× faster than pure-Rust CPU using wgpu Metal. - **Volume-to-Surface Projection**: Load raw NIfTI fMRI and project to cortical surface. - **NIfTI Volume Output**: Surface-to-volume projection with 6mm Gaussian smoothing. - **HCP-MMP1 ROI Analysis**: Per-region brain activation summaries. - **Evaluation Metrics**: Pearson correlation, MSE, top-k retrieval vs ground-truth fMRI. - **Per-Modality Contribution Maps**: Ablation-based text/audio/video contribution. - **Stimulus-Aligned Visualization**: HTML with video frames, brain activity, and text aligned. - **Text-to-Speech**: Text input → TTS audio → whisperX → word events → predictions. - **MP4 Video Output**: Animated brain activity over time via ffmpeg. - **8 Numeric Parity Tests**: Verified identical to Python across all 3 backends. ### Tech Stack: Rust ### Why It Matters: The `tribev2-rs` crate provides a pure-Rust inference engine for the TRIBE v2 model, ensuring the same pretrained weights as the original Hugging Face model. This ensures numerical parity with the Python reference implementation, making it a reliable and efficient choice for inference tasks

### Overview: Nano Banana Pro API (`gemini-3-pro-image-preview`) is a powerful OpenAI-compatible endpoint that allows developers to generate and edit high-quality images at a cost-effective price. Ideal for businesses and individuals looking to integrate image generation into their applications without the need for a monthly subscription. ### Key Features: - **Model ID and Pricing**: The API uses the same base URL and headers as the OpenAI-compatible endpoint, making it easy to integrate with existing systems. - **Pricing Model**: Nano Banana Pro is priced on a per-image basis, starting at $0.03 per image (4K at $0.04). Users can pay as they go without needing a monthly subscription. - **Parameter Flexibility**: Supports various parameters such as `model`, `prompt`, `size`, `resolution`, and `n` to customize image generation. - **Cobranching**: The cost is reflected in the response, and tasks marked as failed are refunded. ### Tech Stack: - Python ### Why It Matters: - **Cost-Effective**: The API offers a cost-effective solution for image generation, with prices starting at $0.03 per image. - **Ease of Use**: The same base URL and headers as OpenAI-compatible endpoints make it easy to integrate with existing systems. - **Customizability**: Offers flexibility in parameters and allows users to pay as they go without the need for a monthly subscription. ### Repository: - Repository: f0xtn9oh8g7ia6/nano-banana-pro-api-pt

Overview: Nano Banana API is a fast, OpenAI-compatible image generation service that allows users to generate images or edit reference images with a single endpoint. Key features: - One endpoint for generating images and editing reference images - Pricing based on the number of images generated, with a minimum charge of $0.0125 per image - Supports multiple image resolutions and editing prompts - Asynchronous processing, with task IDs for tracking progress Tech stack: Python Why it matters: Nano Banana API simplifies the process of generating images or editing reference images, providing a single endpoint for OpenAI-compatible image generation, making it easier for developers to integrate into their projects without the complexity of multiple endpoints. Repository: f84q06w110/nano-banana-api-fr Primary language: Python GitHub stars: 12 GitHub description: Nano Banane / la Banane — Nano Banana API (`gemini-2.5-flash-image-preview`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint. README excerpt:

### Overview: Nano Banana 2 Lite API (`gemini-3.1-flash-lite-image-ext`) is a cost-effective alternative to Nano Banana 2 Lite, offering a single OpenAI-compatible endpoint for generating images at a price of $0.0125 per image. This API supports 300+ models, including Nano Banana 2 Lite, and provides a seamless experience with a single Base-URL and `Authorization` header. ### Key Features: - **Single Endpoint**: A single API endpoint for Nano Banana 2 Lite, supporting 300+ models. - **Pricing**: $0.0125 per image, making it highly cost-effective. - **Model ID**: Supports Nano Banana 2 Lite and 300+ other models. - **Asynchronous**: Tasks are processed asynchronously, ensuring no need for retries or batching. - **Modular Pricing**: Prices are based on the tier of the model, with a minimum of $1 per month. ### Tech Stack: - Python ### Why It Matters: - **Cost Efficiency**: Nano Banana 2 Lite offers a cheaper alternative to Nano Banana 2 Lite, with a per-image pricing model. - **Model Variety**: Supports Nano Banana 2 Lite and over 300 other models, providing flexibility. - **Ease of Use**: Asynchronous processing ensures no need for retries or batching, making it simple to integrate into any project. - **Single API Endpoint**: A single API endpoint for Nano Banana 2 Lite, simplifying the integration process. ### GitHub Stars: -

### Overview: Nano Banana 2 Lite Prompts: 8 Tested Recipes with Outputs ### Key Features: - **Model ID:** `gemini-3.1-flash-lite-image-ext` - **Effective Price:** $0.0125 per image - **Cost per Image:** $0.1 for the entire set ### Tech Stack: Python ### Why It Matters: Nano Banana 2 Lite (`gemini-3.1-flash-lite-image-ext`) provides an open and cost-effective alternative to traditional image generation services, offering a single endpoint that supports multiple models, including 300+ image, video, and speech models. This makes it ideal for developers seeking a versatile and scalable solution without the need for a monthly subscription or prepaid plans. The API is compatible with OpenAI, ensuring seamless integration into existing projects. The API key can be obtained directly from the provided link, and the pricing is straightforward, with each completed task returning the calculated cost. ### Primary Language: Python ### GitHub Stars: 12 ### GitHub Description: Nano Banana 2 Lite prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`gemini-3.1-flash-lite-image-ext`) and the cost per image.

Overview: kibble is a fast, dependency-light toolkit designed to ingest, clean, index, and search data from various sources, including local files, git repos, websites, and Hugging Face datasets. It provides a fast ingestion process and supports fine-tuning and retrieval capabilities. Key features: - Fast ingestion and indexing - Clean and structured datasets - Real-time search capabilities - Integration with external backends for model training and OCR/ASR Tech stack: Rust Why it matters: kibble simplifies the process of acquiring, cleaning, and indexing large volumes of data from diverse sources, making it an essential tool for developers and data scientists looking to automate knowledge acquisition and data management. GitHub stars: 14 Repository: femboyisp/kibble

### Overview: Nano Banana 2 Lite API, developed by the team at APIMart, offers a cost-effective alternative to OpenAI's Gemini 3.1 Flash Lite Image route. This API is priced per image and provides access to Nano Banana 2 Lite and over 300 other image, video, and language models. The API endpoint is `https://api.apimart.ai/v1/images/generations`, and it supports asynchronous processing, making it easy to integrate into applications. ### Key features: - **Pricing**: Nano Banana 2 Lite is priced per image, starting at $0.0125 per image. - **Model ID and Routes**: The API supports the `gemini-3.1-flash-lite-image-ext` model, which can be accessed via the route `https://api.apimart.ai/v1/images/generations`. - **Billing Model**: The pricing model is based on the number of images generated, with a minimum charge of $1 per month. - **OpenAI-Compatible Endpoint**: The API endpoint is compatible with OpenAI's API, allowing for seamless integration with existing OpenAI models. - **Asynchronous Processing**: The API is designed to be asynchrounous, making it easy to integrate into applications without requiring custom integration. ### Tech stack: - **Python**: The API is developed using Python, utilizing the `requests` library for making HTTP requests. ### Why it matters: - **Cost-Effective**: Nano Banana 2 Lite is significantly cheaper than OpenAI's Gemini 3.1 Flash Lite

Overview: This project introduces an Autonomous Coding Agent that leverages Restack, OpenAI's GPT models, Docker-in-Docker, and a frontend React UI to automatically generate, run, and validate code based on user-provided prompts. Key features: - Dockerfile-based Autonomous Coding Agent - RESTack integration for code generation - OpenAI GPT models for code generation and validation - Interactive frontend UI for real-time monitoring and advanced editing Tech stack: TypeScript Why it matters: This project represents a significant advancement in the field of autonomous coding, offering developers a powerful tool to automate the coding process and ensure the quality of their code through continuous validation and feedback. Repository: hem9984/Azlon-demo Primary language: TypeScript GitHub stars: 14 GitHub description: The world's first Dockerfile-based Autonomous Coding Agent

Overview: This project introduces a large semi-automatically generated dataset containing ~400,000 descriptive sentences about commonsense knowledge, focusing on sentences that can be true or false with negation present in about 2/3 of the corpus. The dataset is designed to evaluate Large Language Models (LLMs). Key features: - Large semi-automatically generated dataset - Contains ~400,000 sentences about commonsense knowledge - Negation present in about 2/3 of the corpus - Available in the 🤗HuggingFace Hub - Provides code for training, evaluating, and scoring LLMs Tech stack: Python Why it matters: This dataset serves as a benchmark to challenge and evaluate Large Language Models (LLMs) by providing a comprehensive set of sentences with varying degrees of negation, ensuring a robust assessment of model performance. Repository: hitz-zentroa/This-is-not-a-Dataset Primary language: Python GitHub stars: 14 GitHub description: We introduce a large semi-automatically generated dataset of ~400,000 descriptive sentences about commonsense knowledge that can be true or false in which negation is present in about 2/3 of the corpus in different forms that we use to evaluate LLMs.

Overview: contextsearch is a self-hosted search engine that indexes various document formats, watches for changes, and provides keyword, dense, and hybrid search results with detailed explanations. Key features: - Supports indexing of PDF, DOCX, PPTX, TXT, Markdown, HTML, CSV, and additional text formats. - Watches folders for changes and rescans periodically to catch missed events. - Offers keyword search (BM25), dense search (sentence-transformers models), and hybrid search (both). - Provides reranking with a cross-encoder and explains every result with details like retrievers, cosine similarity, and more. - Compares search strategies side by side and scores them using `contextsearch eval` for recall, MRR, nDCG, and latency. - Runs as a single process with SQLite storage and no external services. - Includes a web UI, REST API, CLI, and JSON logs. Tech stack: Python, SQLite, sentence-transformers, Hugging Face cache. Why it matters: contextsearch is a powerful, self-hosted search engine that simplifies document search, providing detailed explanations and strategies for various search methods, making it ideal for developers and professionals looking for a robust and customizable search solution.

### Overview Prometheus BCI is an adaptive and inclusive neurotech platform designed for real-time multimodal data analysis. It processes inputs from EEG, PPG, ECG, and camera feeds, converting raw data into actionable metrics and insights for cognitive and physiological analysis. ### Key Features 1. **Multimodal Data Collection** - Supports connections with Emotiv, OpenBCI, BITalino, EmotiBit, and cameras. 2. **Signal Processing** - Filters and features extract raw input data. ### Tech Stack Python ### Why It Matters Prometheus BCI is a testament to the power of neurotechnology and the importance of inclusivity in innovative applications. It demonstrates how advanced AI and neurotechnology can be used to support people with disabilities, inspired by the stories of Nathalie and Pierre. GitHub stars: 14 Contributing - Open source, community-driven - Fork, create feature branches, propose changes - Implement unit tests for new features and bug fixes - Document methods and references for new pipelines Feel free to contribute, whether it's advanced Riemannian geometry, CSP, deep learning, or creative signal processing. Together, let's advance open science and BCI innovation!

### Overview: Seedance 1.5 Pro Prompts: 8 Tested Recipes with Outputs Dreamina — Seedance 1.5 Pro prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`seedance-1.5-pro`) and the cost per image. ### Key Features: - **8 Prompts**: Four recipes for ceramic mugs, a cinematic rainy street, a product commercial, and a macro shot. - **Example Renders**: Each recipe includes an example render, duration, and the cost on the APIMart route. - **Model ID**: `seedance-1.5-pro`, compatible with OpenAI's API endpoint. ### Tech Stack: Python ### Why It Matters: - **Cost Efficiency**: The cost per second is $0.0104, making it a cost-effective option for image prompts. - **Versatility**: The model supports 300+ image, video, and language models, ensuring compatibility with various use cases. - **Real-time Pricing**: The cost is calculated and displayed in real-time, providing transparency and control over expenses. ### Repository: j1684-htxd147q/doubao-seedance-1-5-pro-prompts-es ### Primary Language: Python ### GitHub Stars: 14 ### GitHub Description: Dreamina — Seedance 1.5 Pro prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`seedance-1.5-pro`) and the cost per image.

Overview: FLUX 2 Max API is a comprehensive endpoint for OpenAI-compatible image generation, offering pricing per megapixel and flexible options for model selection and image resolution. Key features: - Supports multiple image, video, and language models - Provides a single endpoint for OpenAI-compatible API access - Offers pricing per megapixel, with tiered options for different resolutions - Supports dynamic pricing based on completed tasks - Assists users in obtaining an API key and accessing pricing information Tech stack: Python Why it matters: FLUX 2 Max API simplifies the process of accessing OpenAI-compatible image generation services, providing a single, scalable solution for developers looking to integrate this functionality into their projects without the need for multiple API keys or pre-paid plans. GitHub stars: 14

Overview: EasyLLM is a Python-based framework designed to make it easy to train and fine-tune Large Language Models, including both Large Language Models and Vision Language Models, using minimal resources and low cost. Key features: - Supports GPU and NPU training - Directly trains on tokenized data using PySpark - Flash speed during fine-tuning due to no redundant computation - Utilizes PCIE for high performance at 20 billion model size - Minimalist implementation of Sequence Parallelism for extra long context - High performance in Visual Language Model fine-tuning - Optimizations for RLHF and Mixture of Experts training - Supports both pre-training and fine-tuning - Minimal communication and dynamic experts balancing for efficient training Tech stack: Python Why it matters: EasyLLM simplifies the process of training and fine-tuning Large Language Models, making it accessible and cost-effective for developers and researchers.

### Overview: Seedance 1.0 Pro Prompts: 8 Tested Recipes With Outputs Seedance 1.0 Pro Prompts offers tested image prompt recipes, including example renders, aspect ratios, and the model ID (`seedance-1.0-pro`). This repository provides a comprehensive guide on how to use Seedance 1.0 Pro through APIMart, including pricing, model compatibility, and how to get started quickly. ### Key Features: - **Model ID:** `seedance-1.0-pro` - **Pricing:** $0.0184 per second on APIMart (billed by resolution tier) - **Compatible with:** OpenAI-compatible endpoint ### Tech Stack: - Python ### Why It Matters: - **One Key Access:** The same base URL and authorization header provide access to Seedance 1.0 Pro and 300+ image models. - **Minimal Cost:** No need for a monthly plan or pre-paid rate; start with $1 and use as needed. - **Return on Cost:** Each completed task returns the cost, and failed tasks are billed to the balance. - **Asynchronous Nature:** Sends tasks, gets `task_id`, and queries `GET /v1/tasks/{id}` for package processing and retries. ### Also Known As: - Sidenс - Sиденс - Sиданс - Sидэнс ### Seedance 1.0 Pro Prompts: - **4 Recipes:** Provided in `data/prompts.json` and `PROM

### Overview: Grok Imagine 1.5 Prompts: 8 Tested Recipes with Outputs ### Key Features: - **Model:** `grok-imagine-1.5-apimart` - **Recipes:** 8 (machine-readable in `data/prompts.json`, table in `PROMPTS.md`) - **Effective Price:** $0.015 per image - **Cost:** $0.12 for the entire set ### Tech Stack: Python ### Why It Matters: Teams use Grok Imagine 1.5 (`grok-imagine-1.5-apimart`) with APIMart to access a single OpenAI-compatible endpoint, enabling cost-effective and flexible image generation without the need for monthly subscriptions or prepaid plans. The prompt structure is designed to optimize rendering by focusing on specific elements like material, light, framing, and setting, ensuring clear and effective images.

Overview: ComfyUI-Higgsfield-Direct is a direct integration of the Higgsfield API into ComfyUI, providing unified access to 100+ generative models without the need for a credits system. It supports multi-model access, reference image support, resolution control, and aspect ratio presets, among other features. Key features: - Multi-model access - Reference image support - Resolution control - Aspect ratio presets - Video generation - Auto-save - Hash Vault compatibility - Installation instructions Tech stack: Python Why it matters: This tool simplifies the process of using Higgsfield models in ComfyUI, offering a seamless and efficient way to generate images and videos without the need for external credits or middlemen. Repository: jeremieLouvaert/ComfyUI-Higgsfield-Direct Primary language: Python GitHub stars: 14 GitHub description: Direct Higgsfield API integration for ComfyUI — multi-model image & video generation (Soul 2.0, Seedream 4, Kling 2.1, and more). No credits system, uses your own API key.

Overview: Seedance 1.5 Pro API (`seedance-1.5-pro`) is a model id, pricing, and parameters API for OpenAI-compatible endpoints, providing model id, pricing, parameters, and copy-paste examples. Key features: - Model ID: Seedance 1.5 Pro (`seedance-1.5-pro`) - Pricing: Per generated second, billed by resolution tier - Parameters: Prompt, duration, resolution, size, image URLs, seed Tech stack: Python Why it matters: Seedance 1.5 Pro API offers a seamless integration with OpenAI-compatible endpoints, providing a scalable and cost-effective solution for AI video generation. Repository: jif7hyqckmlct/seedance-1-5-pro-api-pt Primary language: Python GitHub stars: 14 GitHub description: Dreamina — Seedance 1.5 Pro API (`seedance-1.5-pro`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint.

### Overview: 即梦图片 — Seedream 5.0 Flash prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`seedream-5-0-flash`) and the cost per image. ### Key features: - **Model ID:** `seedream-5-0-flash` - **Example Renders:** Eight tested recipes with detailed prompts and example renders. - **Aspect Ratios:** 1:1, 4:3, 16:9, 85mm, 24mm, macro, overhead flat lay. - **Cost:** $0.0137 per image on the APIMart route. ### Tech stack: Python ### Why it matters: - **Key:** A single API key allows access to all models and pricing. - **Cost:** $1 per charge, no need for pre-paid subscriptions. - **Price:** $0.0137 per image, with real-time cost returned on each task. - **Async Design:** Simple queue-based task submission and polling for updates. ### Repository: jikbtcp30/seedream-5-0-flash-prompts-cn ### Primary language: Python ### GitHub stars: 14