Slice

What Slice MCP server does

Slice processes audio and video files for podcast and short-form content workflows. An episode can be uploaded, transcribed, searched for moments worth publishing, and rendered as captioned vertical video. Audio-only sources such as MP3, M4A, WAV, OGG, and FLAC use an audio-based visual template for the resulting clips.

The server also supports transcription as a separate workflow. It returns speaker turns and a results page with TXT, SRT, and VTT downloads. This is useful when the goal is to obtain the spoken words rather than create social clips.

How it works

For clip production, start_episode accepts a file attached in the conversation. Slice copies the file, transcribes it, and searches for candidate moments. If no file is attached, the tool returns an upload URL for the user to open and use. Episodes can be up to three hours long and 4.5 GB. get_episode reports the pipeline state, including upload, transcription, clip discovery, completion, and failure states.

Once an episode is ready, make_clips renders selected candidates. Selection can use the clip numbers or references returned by get_episode. Each render produces a job that can be monitored with get_job, which can return progress, a still image, and a download link when complete.

A completed clip can be rendered again with restyle_clip. Caption style, size, position, hook title, fit, and background can be changed. A preferred look can also be saved as the default for new clips. find_clips_about searches a ready episode again for a topic or audience focus and adds matching candidates without repeating clips already found.

Tools and capabilities

• start_episode: Begin episode processing from an attached file or create an upload URL.
• get_episode: Check processing status, list recent episodes, and inspect discovered clips.
• make_clips: Cut and caption selected clips.
• get_job: Monitor renders and retrieve finished files.
• restyle_clip: Re-render an existing clip with different presentation settings.
• list_styles: Inspect available caption styles and their previews.
• find_clips_about: Search a ready episode for moments matching a subject or audience.
• read_captions: Inspect captions by time range or matching text, including speaker and on-screen wording.
• transcribe_audio: Convert an attached file or direct HTTPS media-file URL to text.
• transcription_status: Check a longer transcription that is still processing.
• email_transcript: Email a completed transcript to the confirmed address on the associated Slice account.

Limitations and notes

Clip creation is designed for uploaded files; YouTube, Spotify, Apple Podcasts, and other webpage links are not accepted for episode processing or transcription. Direct HTTPS links are supported by transcribe_audio only when they point to media files such as MP3, M4A, WAV, or MP4.

start_episode and the clip-search tools require status checks before later operations: clips must be ready before they can be rendered or searched by topic. Transcription accepts files up to 90 minutes. The first transcript does not require an account; later use requires Slice sign-in, and signed-in users receive two transcripts per month shared with the Slice website. Emailing requires sign-in and sends files only to the confirmed account email. Each transcript can be emailed once.