gifhorse

What This Skill Does

GifHorse is a powerful OpenClaw agent skill designed to bridge the gap between vast video libraries and expressive digital communication. It functions by indexing video dialogue, allowing users to query their local media collection as if it were a searchable database. Beyond mere indexing, GifHorse handles the complex technical lifting of video processing—transcribing audio, syncing subtitles, and rendering high-quality GIF output. By leveraging either online subtitle providers or Whisper AI for local speech-to-text, it ensures that every line of dialogue becomes a discoverable asset. Users can preview their selections to ensure precise timing and utilize advanced features like subtitle text substitution for personalized meme creation.

Installation

Installation requires two primary components: the GifHorse skill itself and the necessary FFmpeg-full dependencies for rendering. First, install the skill via the OpenClaw hub command: clawhub install openclaw/skills/skills/coyote-git/gifhorse. Once installed, you must prepare your environment. Navigate to the installation directory (cd ~/gifhorse) and ensure the Python virtual environment is initialized. Because GifHorse relies on local system resources, verify that FFmpeg-full is active. For initial usage, run gifhorse transcribe ~/Movies to populate your local database with your video assets. You can utilize source ~/gifhorse/activate.sh to quickly access the command-line interface for manual batch operations.

Use Cases

GifHorse is ideal for content creators, social media managers, and developers who need to quickly generate reaction media without navigating complex video editing software. Use it to build a personal library of "in-jokes" for internal chat platforms, generate professional clips for technical documentation, or simply create high-quality, perfectly timed memes for social media sharing. Because it supports text substitution, it is particularly effective for recontextualizing famous movie scenes into software development or corporate-themed satire.

Example Prompts

1. "Search my library for the scene from The Lord of the Rings where someone says 'the age of men' and create a 5-second GIF, but change 'men' to 'legacy code'."
2. "Find all instances of 'I have a bad feeling about this' across my library and show me previews so I can pick the best one."
3. "Transcribe my new 'Training Videos' folder and then let me know when you've finished indexing so I can start making clips."

Tips & Limitations

To get the best results, ensure your video library is organized. While Whisper AI is incredibly accurate, it is resource-intensive; prefer using fetch-subtitles first for speed. If you find the subtitles are misaligned, use the --include-before and --include-after flags to capture more context. Note that GifHorse requires significant local disk space if you are transcribing high-resolution media with Whisper, as the database files can grow over time. Always verify your FFmpeg installation if renders appear corrupted or missing frames.