assemblyai-transcribe_skill

Overview

This skill transcribes audio and video using AssemblyAI, supporting local uploads or remote URLs and exporting results as plain text, subtitles (SRT/VTT), paragraphs, or sentences. It provides a small CLI helper to run the standard AssemblyAI REST flow, manage polling, and save outputs to files. The tool is designed for quick integration into automation and archival workflows.

How this skill works

The helper script uploads local files to AssemblyAI (POST /v2/upload) or accepts a public URL, then creates a transcript job (POST /v2/transcript) and polls the transcript endpoint (GET /v2/transcript/:id) until completion. It can then export the transcript in several formats (text, JSON, SRT, VTT, paragraph and sentence groupings) and supports passing any advanced transcript options AssemblyAI accepts.

When to use it

• Transcribe interviews, meetings, podcasts, or archival audio/video for searchable text.
• Generate subtitles (SRT/VTT) to add captions to videos.
• Export structured reads like paragraphs or sentence-level splits for downstream processing.
• Automate bulk transcription in scripts or CI/CD pipelines where CLI control is preferred.
• Require EU data residency by directing requests to AssemblyAI's EU endpoint.

Best practices

• Provide ASSEMBLYAI_API_KEY via environment variables; avoid logging keys or embedding them in code.
• Use --out <file> for long or large transcripts to avoid console buffering and truncation.
• When using uploads, ensure the upload and transcript requests use the same API key/project to prevent 403 errors.
• Pass advanced options through --config for speaker labeling, summarization, or domain models as needed.
• Set ASSEMBLYAI_BASE_URL to the EU host if you need EU processing or residency.

Example use cases

• Archive a podcast episode: transcribe local MP3 and export both text and SRT for publishing.
• Caption a conference recording: transcribe MP4 and export VTT for player integration.
• Bulk process an audio archive: run the CLI in a script that uploads files, polls, and saves JSON exports.
• Prepare data for NLP: export sentences or paragraphs for indexing, search, or training datasets.
• On-demand fetch: retrieve an existing transcript by ID and wait for completion before exporting.