Transcriptor MCP

What samson-art/transcriptor-mcp MCP server does

samson-art/transcriptor-mcp MCP server connects MCP-compatible AI clients to video content across 11 supported platforms. Video tools accept a supported video URL or a plain YouTube ID; playlist transcript retrieval accepts a playlist URL or a watch URL containing a list parameter. Responses include human-readable content for chat and structured data for applications.

The server is suited to transcript-based summarization, subtitle extraction, content research, metadata inspection, chapter navigation, and timestamp-based visual inspection. It can also search YouTube and process selected videos from a playlist.

How it works

The hosted service is available at https://transcriptor.gateway.mcpal.io/mcp. An MCP client sends a tool request with a video, playlist, or search input, and receives text plus typed response data. Long transcript and subtitle results are split into parts. A response can include a next_cursor, which the client passes to retrieve the next segment.

Transcript output is cleaned plain text without timestamps, HTML, or speaker names. Raw subtitle retrieval preserves SRT or VTT content and supports official or automatically generated tracks. For platforms where the original language cannot be identified, the client can first request the available tracks and then select a language and subtitle type.

Setup and configuration

Connect the hosted endpoint as an HTTP MCP server using the URL above. The README provides direct installation flows for Cursor, VS Code, and LM Studio, a Claude Code command using the HTTP transport, and connector instructions for Claude, ChatGPT, and Codex. Other MCP clients can use a server entry with the same URL.

The project can also be self-hosted. A Docker image is available, and the Docker image includes ffmpeg, which is required by frame capture.

Tools and capabilities

samson-art/transcriptor-mcp MCP server exposes these tools:

• get_transcript returns cleaned transcript text, with language selection and pagination support.
• get_raw_subtitles returns paginated SRT or VTT data for an official or auto-generated track.
• get_available_subtitles lists available official and automatic language tracks.
• get_video_info returns identity, uploader, counts, classification, thumbnail, and other metadata.
• get_video_chapters returns chapter titles with start and end times.
• get_video_frame captures a JPEG or PNG at a timecode or number of seconds, with width and JPEG quality options.
• get_playlist_transcripts retrieves transcripts for selected playlist items, with a maximum item count.
• search_videos searches YouTube with result limits, offsets, upload-date filters, and JSON or Markdown output.

Interactive interfaces are available for transcript, metadata, frame, and video-search tools in clients that support MCP Apps or the ChatGPT Apps SDK. Other clients receive the same information as text and JSON.

Limitations and notes

Frame capture requires ffmpeg; the supplied Docker image includes it. Frame width cannot exceed 1920 or the source dimensions, and JPEG quality settings apply only to JPEG output. get_video_frame requires either a timecode or seconds value, not both.

Search is documented for YouTube, while the other video tools accept links from the supported platforms. Playlist language selection is required for batch transcript retrieval. For a single video, the server can select the original language when it can identify it; otherwise, call the subtitle-listing tool first. samson-art/transcriptor-mcp MCP server does not remove the need to handle pagination when transcripts or subtitles exceed one response.