AgentSkillsCN

transcribe-and-analyze

使用WhisperKit在本地对来自URL（YouTube、直接媒体链接）的音频与视频进行转录。若用户明确要求，还可选择性地利用AI对转录内容进行分析。当用户提供了媒体内容的URL，并请求转录或语音转文字时，可使用此技能。

SKILL.md
--- frontmatter
name: transcribe-and-analyze
description: Transcribe audio and video from URLs (YouTube, direct media links) using WhisperKit locally. Optionally analyze transcripts with AI when explicitly requested. Use when users provide URLs to media content and request transcription or speech-to-text conversion.

Transcribe and Analyze

Local transcription of audio/video content using WhisperKit. Analysis is available on request using OpenAI or local Ollama.

Capabilities

  1. •Transcription - Convert audio/video URLs to text using WhisperKit (runs locally, always available)
  2. •Analysis - Extract insights from transcripts (only when user asks for it, supports OpenAI or Ollama)

Quick Start

Transcribe Only

bash
python3 scripts/transcribe.py "https://youtube.com/watch?v=..."

Transcribe + Analyze (OpenAI)

bash
python3 scripts/transcribe.py "https://youtube.com/watch?v=..."
python3 scripts/analyze_transcript.py whisper-transcriptions/video.md

Transcribe + Analyze (Local)

bash
python3 scripts/transcribe.py "https://youtube.com/watch?v=..."
python3 scripts/analyze_transcript.py whisper-transcriptions/video.md --local

Transcription

Script Options

bash
# Basic
python3 scripts/transcribe.py "URL"

# Custom output directory
python3 scripts/transcribe.py "URL" --output-dir "/path/to/save"

# Higher accuracy (slower)
python3 scripts/transcribe.py "URL" --model medium

# Without timestamps
python3 scripts/transcribe.py "URL" --no-timestamps

# Custom filename
python3 scripts/transcribe.py "URL" --filename "my-transcription.md"

Whisper Models

ModelSpeedAccuracyUse Case
tinyFastestLowestQuick drafts, testing
baseFastReasonableSimple content
smallBalancedGoodDefault - most use cases
mediumSlowerHighLectures, important content
largeSlowestHighestCritical accuracy needed

Dependencies

Script checks for these and provides install instructions if missing.

Output

Transcriptions save to ./whisper-transcriptions/ as markdown:

markdown
# Transcription

**Source:** https://youtube.com/watch?v=example
**Transcribed:** 2025-01-15 14:30:00
**Tool:** WhisperKit

---

[00:00:00.000 --> 00:00:05.000] Welcome to this video...

Analysis

Provider Options

OpenAI API (default):

bash
python3 scripts/analyze_transcript.py transcript.md
python3 scripts/analyze_transcript.py transcript.md --model gpt-4o

Requires OPENAI_API_KEY environment variable.

Local Ollama:

bash
python3 scripts/analyze_transcript.py transcript.md --local
python3 scripts/analyze_transcript.py transcript.md --local --model mistral

Requires Ollama running (ollama serve).

Script Options

bash
# Default comprehensive analysis (OpenAI)
python3 scripts/analyze_transcript.py transcript.md

# Use local Ollama
python3 scripts/analyze_transcript.py transcript.md --local

# Specify model
python3 scripts/analyze_transcript.py transcript.md --model gpt-4o
python3 scripts/analyze_transcript.py transcript.md --local --model llama3.2

# Custom analysis prompt
python3 scripts/analyze_transcript.py transcript.md --prompt "List all tools mentioned"

# Custom output location
python3 scripts/analyze_transcript.py transcript.md --output ~/Documents/analysis.md

# Print to stdout instead of saving
python3 scripts/analyze_transcript.py transcript.md --print

Default Analysis Includes

  • •Executive summary (2-3 paragraphs)
  • •Key insights (5-7 bullet points)
  • •Topics discussed with summaries
  • •Notable quotes (3-5 memorable quotes)
  • •Action items and recommendations
  • •Additional observations

Custom Prompt Examples

bash
--prompt "List all technologies and tools mentioned"
--prompt "What are the main arguments presented?"
--prompt "Extract all statistics and data points"
--prompt "Summarize in 5 bullet points"
--prompt "What questions were asked and how were they answered?"

Dependencies

  • •openai - pip install openai (used for both OpenAI and Ollama)
  • •OPENAI_API_KEY environment variable (for OpenAI only)
  • •Ollama running locally (for --local mode)

Output

Analysis saves alongside transcript as transcript_name_analysis.md:

markdown
# Transcript Analysis

**Source Transcript:** path/to/transcript.md
**Analysis Model:** gpt-4o-mini (OpenAI)
**Tokens Used:** 33,763

---

[Analysis content]

Common Workflows

Full Pipeline: URL to Insights (Cloud)

bash
python3 scripts/transcribe.py "https://youtube.com/watch?v=abc123"
python3 scripts/analyze_transcript.py whisper-transcriptions/watch.md

Full Pipeline: URL to Insights (Local)

bash
python3 scripts/transcribe.py "https://youtube.com/watch?v=abc123"
python3 scripts/analyze_transcript.py whisper-transcriptions/watch.md --local

Multiple Analyses on Same Transcript

bash
python3 scripts/analyze_transcript.py transcript.md --output summary.md
python3 scripts/analyze_transcript.py transcript.md --prompt "List action items" --output actions.md
python3 scripts/analyze_transcript.py transcript.md --prompt "Extract quotes" --output quotes.md

Batch Transcription

bash
python3 scripts/transcribe.py "URL1"
python3 scripts/transcribe.py "URL2"
python3 scripts/transcribe.py "URL3"

Reference Files

Troubleshooting (references/troubleshooting.md)

  • •Download failures
  • •Transcription errors
  • •Dependency issues
  • •API errors

Configuration (references/configuration.md)

  • •Output format details
  • •File naming behavior
  • •Model selection guidance

Usage Patterns (references/usage-patterns.md)

  • •Common transcription scenarios
  • •Analysis patterns
  • •Batch processing tips

Scripts

ScriptPurpose
scripts/transcribe.pyDownload and transcribe audio/video from URLs (WhisperKit)
scripts/analyze_transcript.pyAI analysis of transcript files (OpenAI or Ollama)