Files

155 lines
4.1 KiB
Bash

#!/usr/bin/env bash
set -euo pipefail
# GETTR Transcribe Pipeline
# Usage: run_pipeline.sh [--language <code>] <gettr_post_url> [output_base_dir]
#
# Runs the full pipeline:
# 1. Extract video URL and slug from GETTR post
# 2. Download audio (16kHz mono WAV)
# 3. Transcribe with MLX Whisper (VTT output)
#
# Output: <output_base_dir>/gettr-transcribe-summarize/<slug>/
# - audio.wav
# - audio.vtt
#
# The summary step (Step 4) is left to the LLM.
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
# Parse optional --language flag
LANGUAGE=""
while [[ $# -gt 0 ]]; do
case "$1" in
--language|-l)
if [[ -n "${2:-}" ]]; then
LANGUAGE="$2"
shift 2
else
echo "[error] --language requires a language code (e.g., zh, en, ja)" >&2
exit 2
fi
;;
-*)
echo "[error] Unknown option: $1" >&2
exit 2
;;
*)
break
;;
esac
done
if [[ $# -lt 1 ]]; then
echo "Usage: run_pipeline.sh [--language <code>] <gettr_post_url> [output_base_dir]" >&2
echo "" >&2
echo "Options:" >&2
echo " --language, -l Language code for transcription (e.g., zh, en, ja, ko)" >&2
echo " If not specified, language is auto-detected." >&2
echo " output_base_dir Base directory for output (default: ./out)" >&2
echo "" >&2
echo "Common language codes:" >&2
echo " zh = Chinese, en = English, ja = Japanese, ko = Korean" >&2
echo " es = Spanish, fr = French, de = German, ru = Russian" >&2
exit 2
fi
GETTR_URL="$1"
OUTPUT_BASE="${2:-./out}"
# Check prerequisites
for cmd in python3 ffmpeg mlx_whisper; do
if ! command -v "$cmd" >/dev/null 2>&1; then
echo "[error] Required command not found: $cmd" >&2
if [[ "$cmd" == "ffmpeg" ]]; then
echo "[hint] Install with: brew install ffmpeg" >&2
elif [[ "$cmd" == "mlx_whisper" ]]; then
echo "[hint] Install with: pip install mlx-whisper" >&2
fi
exit 127
fi
done
echo "=== Step 1: Extracting video URL and slug ===" >&2
# Run extraction script and capture output
EXTRACT_OUTPUT=$("$SCRIPT_DIR/extract_gettr_og_video.py" "$GETTR_URL") || {
echo "[error] Failed to extract video URL from GETTR post" >&2
exit 1
}
# Parse output: line 1 = video URL, line 2 = slug
VIDEO_URL=$(echo "$EXTRACT_OUTPUT" | head -n1)
SLUG=$(echo "$EXTRACT_OUTPUT" | tail -n1)
if [[ -z "$VIDEO_URL" || -z "$SLUG" ]]; then
echo "[error] Failed to parse extraction output" >&2
exit 1
fi
echo "[info] Video URL: $VIDEO_URL" >&2
echo "[info] Slug: $SLUG" >&2
# Set up output directory
OUT_DIR="$OUTPUT_BASE/gettr-transcribe-summarize/$SLUG"
mkdir -p "$OUT_DIR"
AUDIO_FILE="$OUT_DIR/audio.wav"
VTT_FILE="$OUT_DIR/audio.vtt"
echo "" >&2
echo "=== Step 2: Downloading audio ===" >&2
"$SCRIPT_DIR/download_audio.sh" "$VIDEO_URL" "$AUDIO_FILE" >&2 || {
echo "[error] Failed to download audio" >&2
exit 1
}
echo "[info] Audio saved: $AUDIO_FILE" >&2
echo "" >&2
echo "=== Step 3: Transcribing with MLX Whisper ===" >&2
# MLX Whisper transcription with optimized flags:
# -f vtt: VTT format for timestamps
# --condition-on-previous-text False: prevents hallucination propagation
# --word-timestamps True: more precise timing (if supported)
# --language: explicit language if provided (otherwise auto-detected)
# Build language flag if specified
LANG_FLAG=""
if [[ -n "$LANGUAGE" ]]; then
LANG_FLAG="--language $LANGUAGE"
echo "[info] Using explicit language: $LANGUAGE" >&2
else
echo "[info] Language will be auto-detected" >&2
fi
mlx_whisper "$AUDIO_FILE" \
-f vtt \
-o "$OUT_DIR" \
--model mlx-community/whisper-large-v3-turbo \
--condition-on-previous-text False \
--word-timestamps True \
$LANG_FLAG \
2>&1 || {
echo "[warn] Retrying without extra flags..." >&2
mlx_whisper "$AUDIO_FILE" \
-f vtt \
-o "$OUT_DIR" \
--model mlx-community/whisper-large-v3-turbo \
$LANG_FLAG
}
echo "[info] Transcript saved: $VTT_FILE" >&2
echo "" >&2
echo "=== Pipeline complete ===" >&2
echo "[info] Output directory: $OUT_DIR" >&2
echo "" >&2
# Output the paths for programmatic use (stdout)
echo "$OUT_DIR"
echo "$AUDIO_FILE"
echo "$VTT_FILE"