Choose theme
AI Resources
Claude Video
Claude Video is an agent skill for asking questions about video URLs or local video files from supported local coding-agent hosts.
Its local engine prepares timestamped frames and transcript material; a Gemini engine sends video for direct understanding. The auto mode selects Gemini when its key is available, so processing depends on the engine and transcription route chosen. Use this as a first read, not a recommendation. Open the original project before trusting details like terms, limits, privacy, cost, setup, or safety.
What it is
A video-input skill for agents
Claude Video packages a SKILL.md file and Python scripts for video questions. Its local engine uses captions or transcription alongside ffmpeg-selected frames; Gemini is a separate cloud video-analysis path.
Why it stands out
Frames and words stay on one timeline
With the local engine, timestamped transcript segments and selected frames share one timeline. Detail modes, focused time ranges and frame deduplication control how much visual material reaches the agent. Explicit timestamps can request frames from particular moments.
Availability
Public skill with several install paths
The current README supports Claude Code, including Desktop Code sessions, Codex and other documented local skill hosts. It explicitly excludes Claude Chat at claude.ai, uploading watch.skill there, and Cowork.
Why it matters
What makes it useful
A transcript can explain what was said while missing a broken interface, a chart change, a visual joke, or the exact moment a demonstration stops working. Claude Video gives an agent both streams, which makes screen-recording diagnosis, visual walkthroughs, and questions about a specific moment more practical.
What to know
Where it fits
Use it when a coding agent needs to inspect a screen recording, tutorial, presentation, product video, or other clip where the picture matters alongside the words. Focused start and end times are especially useful when the question concerns one part of a long video.
Notable points
What stands out
Version 0.3.2 removes claude.ai and Cowork support. Current transcription choices include local WhisperX, cloud services or captions-only, alongside the separate Gemini engine. Local transcription on a remote host still runs on that host, not on the reader’s device.
Before using
What to review
Whether the supported host can run Python and shell commands, and whether yt-dlp, ffmpeg and the selected transcription dependencies are acceptable to install.
Whether the video source's platform rules and recording rights allow the intended download and use.
Which engine and transcription route will process the material. Gemini receives video; cloud transcription receives audio. Local transcription still leaves the surrounding agent’s model and tool connections to check.
How much context the local engine’s frame mode consumes. Capped modes become sparse on recordings longer than about 10 minutes; focused ranges can keep the request relevant.
Where temporary frames, audio, captions and transcripts are created, what the agent can read and when those files should be removed.
The local engine uses available captions first. Decide whether those native captions are sufficient before choosing a transcription route.
For an exact-moment question, --timestamps 1:05,2:30 requests cue frames at those times, including in transcript-only mode. Requested cue frames consume the frame budget first, and the source video still needs to be obtainable.
Reader fit
Who may find it relevant
Users of supported local coding-agent hosts who need to inspect a recording as part of a task.
Builders reviewing screen recordings, tutorials, presentations or time-specific visual evidence.
Less relevant for readers who want a finished video editor, a generative video model or a claude.ai upload.
Editorial note
Why LifeHubber lists it
When a transcript points to a fleeting on-screen event, Claude Video's local engine can request cue frames at chosen timestamps instead of leaving that moment to interval or scene sampling. For a captionless recording you intend to transcribe on the local host, select --engine local and the documented WhisperX route explicitly. Auto mode can choose Gemini when its key is available. Check the local setup first: initial installation downloads packages and models, while a transcription failure does not silently switch to a cloud service.
Source links
Source materials
Reader note
Before relying on this entry
LifeHubber lists entries to help readers inspect AI projects, not to endorse them or prove they are safe, suitable, accurate, maintained, or right for a specific use. We do not verify every entry in depth. Before relying on anything listed, review the original materials, terms, privacy practices, limits, and risks that matter for your situation.
What to explore next
Choose the next layer around the video.
Claude Video handles a question about one recording. Continue with a wider source-access layer, a realtime video-language model, or a local transcription utility when the workflow needs a different boundary.
More in AI Agents
Keep browsing this category
Explore more AI agent projects.
Paperclip
paperclipai/paperclip
A self-hosted server and dashboard for coordinating agent teams through companies, goals, roles, issues, heartbeats, budgets, approvals, and persistent activity records.
AgentMemory
rohitg00/agentmemory
A self-hosted coding-agent memory server with session capture, keyword recall, optional local or remote embeddings, a viewer, and integrations through hooks, MCP, and REST.
Mem0
mem0ai/mem0
A memory layer for AI agents and assistants, with library, self-hosted, platform, SDK, CLI, cookbook, evaluation, and integration paths for persistent context.
For project maintainers
Listed here? You can use the badge.
If you maintain a project with a current LifeHubber listing, you may add the optional “Listed on LifeHubber AI Resources” badge to its README, docs, or website. No introduction or permission request is needed.
