Genr Kit

The open-source playground for generative AI. Rapidly prototype and deploy multi-modal applications—from text and image classification to speech synthesis—using Python, Gradio, and Transformers.

Features

• 🚀 Gradio Interface: Fast, modern AI-Powered apps interface with the API ready feature
• 🤗 Hugging Face Integration: Uses state-of-the-art BLIP model for image captioning
• 📦 Batch Processing: Support for processing multiple images at once
• 🔥 GPU Support: Automatic GPU acceleration when available

Tools

Genr-Kit provides a comprehensive suite of tools for common generative AI tasks. Each tool leverages state-of-the-art, publicly available models from the Hugging Face Hub.

---

📝 Text & NLP Tools

Text Classification

Categorize text into predefined labels like sentiment or topic.

• Model: distilbert-base-uncased-finetuned-sst-2-english A fast and accurate model fine-tuned for sentiment analysis (positive/negative), perfect for real-time applications.

Text Generation

Generate coherent and contextually relevant text from a prompt.

• Model: gpt2 A pioneering transformer model capable of generating creative text continuations in various styles.

Sentiment Analysis

Analyze text to determine its emotional tone (e.g., positive, negative, neutral).

• Model: cardiffnlp/twitter-roberta-base-sentiment-latest A robust model trained on a large corpus of tweets, excellent for modern, informal language.

Text Summarization

Condense long articles, reports, or documents into concise summaries.

• Model: facebook/bart-large-cnn The BART model fine-tuned on the CNN/DailyMail dataset, making it excellent for abstractive summarization.

Named Entity Recognition (NER)

Identify and extract entities like names, organizations, and locations from text.

• Model: dslim/bert-base-NER A BERT model specifically fine-tuned to recognize common named entities with high accuracy.

Text-to-Command

Convert natural language instructions into structured commands or API calls.

• Model: microsoft/DialoGPT-medium While often used for chat, its fine-tuning capabilities make it a good base for learning instruction-to-command tasks.

---

🖼️ Computer Vision Tools

Image Classification

Identify the main subject or scene within an image.

• Model: google/vit-base-patch16-224 A Vision Transformer (ViT) model that achieves excellent accuracy on the ImageNet-1k benchmark.

Object Detection

Locate and identify multiple objects within an image using bounding boxes.

• Model: hustvl/yolos-tiny A You Only Look Once (YOLO) transformer model that provides a great balance of speed and accuracy for real-time detection.

Image Generation

Generate novel images from a text description.

• Model: runwayml/stable-diffusion-v1-5 A powerful latent diffusion model for creating high-quality, detailed images from any text prompt.

Image Captioning

Generate a descriptive English-language caption for a given image.

• Model: Salesforce/blip-image-captioning-base The BLIP model provides high-quality, context-aware captions, ideal for accessibility and content description.

Image-to-Image Translation

Transform an input image based on a text prompt (e.g., style transfer, enhancement).

• Model: timbrooks/instruct-pix2pix A model specifically fine-tuned for following instructions to edit images, like "make it a cartoon".

Image Segmentation

Identify and map specific objects or regions in an image at the pixel level.

• Model: facebook/detr-resnet-50-panoptic A transformer-based model that performs both instance and panoptic segmentation in a single architecture.

---

🔊 Audio & Speech Tools

Text-to-Speech

Convert written text into natural-sounding spoken audio.

• Model: espnet/kan-bayashi_ljspeech_vits A VITS-based model that produces highly natural and expressive speech in English.

Speech-to-Text

Transcribe spoken audio from various languages into written text.

• Model: openai/whisper-base OpenAI's Whisper model offers robust, accurate transcription and translation across multiple languages.

Speech Enhancement

Remove background noise and improve the clarity of an audio recording.

• Model: microsoft/speechbrain-mtl-mimic-voicebank A model trained specifically for speech enhancement and denoising tasks.

Music Generation

Generate short musical audio clips from text descriptions.

• Model: facebook/musicgen-small A simple and controllable model for generating high-quality music from text prompts.

---

🔗 Multi-Modal Tools

Visual Question Answering

Answer natural language questions about the contents of an image.

• Model: dandelin/vilt-b32-finetuned-vqa A vision-and-language transformer (ViLT) model designed for answering questions about images.

Document Question Answering

Answer questions based on the content of a document (e.g., scanned PDFs, images with text).

• Model: impira/layoutlm-document-qa A model that understands the layout of documents (text + spatial information) to answer questions accurately.

Embeddings Generator

Create numerical vector representations (embeddings) of text, images, or audio for analysis and search.

• Model: sentence-transformers/all-MiniLM-L6-v2 A versatile model that maps sentences and paragraphs to a dense vector space, perfect for semantic search and clustering.

Quick Start

Installation

1. Clone the repository:

git clone https://github.com/alinrajpoot/genr-kit.git
cd genr-kit

2. Install dependencies:

pip install -r requirements.txt

3. Run the server:

python main.py

The API will be available at http://localhost:9000

Usage

Once the server is running, visit http://localhost:9000 in browser.

Requirements

• Python 3.8+
• Gradio
• Hugging Face Transformers
• PyTorch
• Pillow (PIL)

License

Open source - feel free to contribute and improve!