Oni Eyes: A Multimodal Social Media Content Analysis Tool Based on OpenRouter

This article introduces the open-source project Oni Eyes (Ghost Eyes), a multimodal media analysis tool designed specifically for social media content analysis. It leverages advanced vision models via the OpenRouter platform to address the automatic understanding of massive social media visual content, suitable for scenarios like brand monitoring, content moderation, research analysis, and more.

Social media generates massive amounts of image and video content daily, but traditional text analysis tools cannot capture visual information, and manual moderation struggles to handle the scale. While Multimodal Large Language Models (MLLMs) can process text and images, developers face challenges in convenient invocation and integration.

Oni Eyes is an open-source multimodal media analysis tool designed for social media, using OpenRouter to call vision models. The name is inspired by the "Oni" (ghost/demon) in Japanese folklore, symbolizing its ability to gain deep insight into visual content. The project is maintained by OhEddie, with source code on GitHub (https://github.com/OhEddie/oni-eyes), released on May 27, 2026.

1. OpenRouter Unified Access Layer: Supports models like GPT-4 Vision, Claude3, Gemini Pro Vision, enabling flexible switching, cost optimization, high availability, and simplified integration. 2. Social Media Adaptation: Tailored to the characteristics of platforms like Instagram/Twitter/TikTok, supports multiple formats, extracts metadata, and identifies trends. 3. Multimodal Capabilities: Scene understanding, OCR text extraction, sentiment analysis, content classification, and safety detection.

1. Social Media Monitoring: Analyze brand UGC, monitor visual assets, identify trends, and track competitor strategies. 2. Content Moderation: Automated violation content identification, sensitive information detection, and batch processing. 3. Research Analysis: Large-scale visual data analysis, automatic annotation and classification, and trend time-series analysis. 4. Personal Management: Organize screenshots, extract information, smart album management, and content strategy analysis.

1. Stream Processing Architecture: Real-time analysis of new content, batch processing of historical data, easy scalability, and resource efficiency. 2. Configurable Strategies: Customizable analysis dimensions, keywords, model sensitivity, and output formats. 3. Privacy and Security: Data minimization, local processing options, encrypted storage and transmission, and compliance with data regulations.

Oni Eyes demonstrates the practical application of multimodal AI in the social media field, and its open-source nature promotes community innovation (such as expanding platforms, integrating workflows, etc.). As social media visual content grows, such tools will become more valuable, providing an example for the popularization of multimodal AI.