Transcript video shell script. Use chunks for longer videos and local faster-wisper. Co-authored with Claude.

transcribe.sh

#!/usr/bin/env bash
set -euo pipefail

if [ $# -lt 1 ]; then
  echo "Uso: $0 ruta/al/video.mp4"
  exit 1
fi

VIDEO_PATH="$1"
if [ ! -f "$VIDEO_PATH" ]; then
  echo "Archivo no encontrado: $VIDEO_PATH"
  exit 1
fi

BASENAME="$(basename "$VIDEO_PATH" .mp4)"
OUT_DIR="$(pwd)/salida_${BASENAME}"
mkdir -p "$OUT_DIR"

AUDIO_PATH="${OUT_DIR}/${BASENAME}.wav"
FINAL_OUTPUT="${OUT_DIR}/${BASENAME}_transcripcion_completa.txt"

echo "=== 1) Extrayendo audio con ffmpeg ==="
ffmpeg -y -i "$VIDEO_PATH" -ac 1 -ar 16000 "$AUDIO_PATH"

echo "=== 2) Dividiendo audio en chunks de 10 minutos ==="
CHUNK_DURATION=600  # 10 minutos en segundos
ffmpeg -i "$AUDIO_PATH" -f segment -segment_time $CHUNK_DURATION -c copy "${OUT_DIR}/chunk_%03d.wav"

echo "=== 3) Transcribiendo cada chunk ==="
rm -f "$FINAL_OUTPUT"  # Limpia output anterior

for chunk in "${OUT_DIR}"/chunk_*.wav; do
  chunk_name=$(basename "$chunk" .wav)
  echo "Procesando $chunk_name..."
  
  faster-whisper "$chunk" \
    --model_size_or_path large-v2 \
    --language es \
    --device cpu \
    --compute_type int8 \
    --cpu_threads 4 \
    -o "${OUT_DIR}/${chunk_name}.txt"
  
  # Añadir al archivo final
  if [ -f "${OUT_DIR}/${chunk_name}.txt" ]; then
    cat "${OUT_DIR}/${chunk_name}.txt" >> "$FINAL_OUTPUT"
    echo "" >> "$FINAL_OUTPUT"  # Salto de línea entre chunks
  fi
done

echo "=== 4) Limpiando archivos temporales ==="
rm -f "${OUT_DIR}"/chunk_*.wav

echo "=== 5) Listo ==="
echo "Transcripción completa en: $FINAL_OUTPUT"
ls -lh "$FINAL_OUTPUT"