Pull a transcript from YouTube or fall back to local transcription.

ytt

#!/bin/bash

# YouTube Transcript Tool (ytt)
# Extracts transcripts from YouTube videos using built-in transcripts or local transcription

url="$1"

if [[ -z "$url" ]]; then
    echo "Usage: ytt <youtube-url>"
    echo "Extract transcript from YouTube video using built-in transcripts or whisperkit-cli fallback"
    exit 1
fi

# echo "$url"

# Try using built-in transcript first
transcript_output=$(yt --transcript "$url" 2>&1)
if echo "$transcript_output" | grep -q "Transcript not available"; then
    echo "[ytt] Transcript not available from YouTube. Falling back to whisperkit-cli transcription..." >&2

    # Temporary directory
    tmp_dir=$(mktemp -d)
    video_file="$tmp_dir/video.mp4"

    # Download video with yt-dlp
    echo "[ytt] Downloading video..." >&2
    if ! yt-dlp -f 'mp4' -o "$video_file" "$url" >/dev/null 2>/dev/null; then
        echo "[ytt] Failed to download video." >&2
        rm -rf "$tmp_dir"
        exit 1
    fi

    # Extract audio from video using ffmpeg
    echo "[ytt] Extracting audio..." >&2
    audio_file="$tmp_dir/audio.wav"
    if ! ffmpeg -i "$video_file" -vn -acodec pcm_s16le -ar 16000 -ac 1 "$audio_file" -y 2>/dev/null; then
        echo "[ytt] Failed to extract audio from video." >&2
        rm -rf "$tmp_dir"
        exit 1
    fi

    # Transcribe with whisperkit-cli (optimized for Apple Silicon)
    if command -v whisperkit-cli &> /dev/null; then
        echo "[ytt] Transcribing with whisperkit-cli..." >&2
        transcript_output=$(whisperkit-cli transcribe --audio-path "$audio_file" --model small --language en --task transcribe 2>/dev/null)
        if [ -n "$transcript_output" ]; then
            echo "$transcript_output"
        else
            echo "[ytt] Failed to generate transcript with whisperkit-cli."
        fi
    elif command -v whisper &> /dev/null; then
        echo "[ytt] Falling back to whisper..." >&2
        cd "$tmp_dir" && whisper "$audio_file" --language en --task transcribe --model small --output_dir . --output_format txt 2>/dev/null
        transcript_txt="$tmp_dir/audio.txt"
        if [ -f "$transcript_txt" ]; then
            cat "$transcript_txt"
        else
            echo "[ytt] Failed to generate transcript with whisper."
        fi
    else
        echo "[ytt] No transcription tool found (whisperkit-cli or whisper)."
        rm -rf "$tmp_dir"
        exit 1
    fi

    # Cleanup
    rm -rf "$tmp_dir"
else
    echo "$transcript_output"
fi