llava

Analyze images and generate grounded answers using CLIP and LLaMA-style models.

Updated May 14, 2026

Search Tags:#fine-tuning #multimodal #vision-language #llava #visual-question-answering #image-chat

Ready to run

npx skills add https://github.com/SethyPagna/Secretary-Jarvis --skill llava-sethypagna

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: llava
Source: https://github.com/SethyPagna/Secretary-Jarvis/tree/main/src/capabilities/optional-skills/mlops/llava
Command: npx skills add https://github.com/SethyPagna/Secretary-Jarvis --skill llava-sethypagna

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

LLaVA solves the gap between image understanding and natural language interaction by letting you ask questions about images, generate descriptions, and carry on multi-turn visual conversations in one workflow.

Core Features & Use Cases

Visual Question Answering: Ask what is in an image, count objects, or inspect a scene with natural language.
Image Captioning and Document Understanding: Turn photos, screenshots, and document images into detailed explanations.
Model Loading and Customization: Work with multiple LLaVA model sizes, quantization options, and fine-tuning paths for research or product prototypes.
Use Case: A product team can upload a screenshot of a dashboard and ask for a summary of charts, labels, and anomalies without building a separate vision pipeline.

Quick Start

Ask LLaVA to describe the attached image and answer a specific question about what it contains.

Frequently Asked Questions about llava

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I build a chatbot that answers questions about images?▼

To build a chatbot for visual question answering, you need a multimodal vision-language architecture that connects a CLIP vision encoder with a Vicuna or LLaMA-style language model to process images and generate grounded text responses.

Can I generate image captions and understand documents from screenshots?▼

Image captioning and document understanding use multimodal models to turn photos, screenshots, and document images into detailed natural language explanations by analyzing the visual content directly.

How do I fine-tune a vision-language model for a custom product prototype?▼

Fine-tuning a vision-language model for product prototypes requires applying quantization options and training paths to adapt the multimodal architecture to your specific visual question answering or image chat dataset.

Does multimodal image chat work for multi-turn conversations about a scene?▼

Multi-turn multimodal chat works by maintaining visual context from an initial image while answering sequential natural language questions, allowing users to inspect a scene and count objects across a conversation.

What components are required to implement visual question answering?▼

Implementing visual question answering requires a CLIP vision encoder, a Vicuna or LLaMA-style language model, and image preprocessing to translate visual content into grounded natural language answers.

What is the best way to summarize charts and labels from a dashboard screenshot?▼

Summarizing dashboard screenshots is best handled by multimodal vision-language models that analyze the image and generate explanations of charts, labels, and anomalies without needing a separate vision pipeline.