now also supports: --output_format (flag)

main.sh

@@ -1,76 +1,114 @@
-# --- args ---
+#!/usr/bin/env bash
-if [[ $# -lt 1 ]]; then
+set -euo pipefail
-  echo "Usage: $0 <file.mp3|file.mp4|URL> [--language <code>] [--model <name>]" >&2
-  exit 1
-fi
 
-WHISPER_EXEC="$(find ~/ -type f -perm -u=x -name whisper)"
+# -------- Defaults --------
-
-INPUT="$1"; shift
 LANGUAGE="en"
 MODEL="medium"
+OUTPUT_FORMAT="txt"     # txt|vtt|srt|json|tsv|lrc|all
+WHISPER_EXEC=""
 
-# Parse flags
+# -------- Usage --------
+usage() {
+  cat >&2 <<EOF
+Usage:
+  $0 <file.mp3|file.mp4|URL> [--language <code>] [--model <name>] [--output_format <fmt>]
+
+Options:
+  --language <code>       ISO code (default: ${LANGUAGE})
+  --model <name>          whisper model name (default: ${MODEL})
+  --output_format <fmt>   One of: txt|vtt|srt|json|tsv|lrc|all (default: ${OUTPUT_FORMAT})
+  -h, --help              Show this help
+
+Notes:
+  - Input may be a local .mp3/.mp4 file or an http(s) URL to .mp3/.mp4
+  - .mp4 will be transcoded to .mp3 via ffmpeg
+EOF
+}
+
+# -------- Arg check --------
+[[ $# -lt 1 ]] && { usage; exit 1; }
+
+INPUT="$1"; shift
+
+# -------- Parse flags --------
 while [[ $# -gt 0 ]]; do
   case "$1" in
-    --language) LANGUAGE="$2"; shift 2 ;;
+    --language)       LANGUAGE="${2:-}"; shift 2 ;;
-    --model)    MODEL="$2";    shift 2 ;;
+    --model)          MODEL="${2:-}";    shift 2 ;;
-    *) echo "Unknown option: $1" >&2; exit 1 ;;
+    --output_format)  OUTPUT_FORMAT="${2:-}"; shift 2 ;;
+    -h|--help)        usage; exit 0 ;;
+    *) echo "Unknown option: $1" >&2; usage; exit 1 ;;
   esac
 done
 
-# --- helpers ---
+# -------- Validate deps --------
+need() { command -v "$1" >/dev/null 2>&1 || { echo "Missing dependency: $1" >&2; exit 1; }; }
+need ffmpeg
+need wget
+
+# Prefer PATH whisper; fallback to $HOME scan
+if command -v whisper >/dev/null 2>&1; then
+  WHISPER_EXEC="$(command -v whisper)"
+else
+  WHISPER_EXEC="$(find "$HOME" -type f -perm -u=x -name whisper 2>/dev/null | head -n1 || true)"
+fi
+[[ -n "$WHISPER_EXEC" ]] || { echo "Could not find 'whisper' executable in PATH or \$HOME" >&2; exit 1; }
+
+# -------- Helpers --------
 lower_ext() {
   local f="$1"
-  local p="${f%%\?*}"             # strip query params if any
+  local p="${f%%\?*}"                 # strip query
-  p="${p%%\#*}"                    # strip fragments
+  p="${p%%\#*}"                        # strip fragment
-  local e="${p##*.}"               # extension
+  local e="${p##*.}"                   # extension
-  printf '%s' "${e,,}"             # lowercase
+  printf '%s' "${e,,}"                 # lowercase
 }
 
-is_url() {
+is_url() { [[ "$1" =~ ^https?:// ]]; }
-  [[ "$1" =~ ^https?:// ]]
+
+output_flags_for() {
+  case "$1" in
+    txt)  echo "--output-txt" ;;
+    vtt)  echo "--output-vtt" ;;
+    srt)  echo "--output-srt" ;;
+    json) echo "--output-json" ;;
+    tsv)  echo "--output-csv" ;;    # whisper.cpp uses CSV/TSV-ish; adjust if your build differs
+    lrc)  echo "--output-lrc" ;;
+    all)  echo "--output-txt --output-vtt --output-srt --output-json --output-csv --output-lrc" ;;
+    *)    echo "Invalid --output_format: $1. Use txt|vtt|srt|json|tsv|lrc|all" >&2; exit 1 ;;
+  esac
 }
 
-# --- temp workspace & cleanup ---
+# -------- Temp workspace & cleanup --------
 WORKDIR="$(mktemp -d)"
 trap 'rm -rf "$WORKDIR"' EXIT
 
-SOURCE_PATH=""   # where the audio/video comes from (local file after download if URL)
+SOURCE_PATH=""
-MP3_PATH=""      # final mp3 we'll feed to whisper
+MP3_PATH=""
 
-# --- stage input (download if URL) ---
+# -------- Stage input (download if URL) --------
 if is_url "$INPUT"; then
-  # Decide ext from URL (we only promise mp3/mp4 support as requested)
   ext="$(lower_ext "$INPUT")"
   case "$ext" in
     mp3|mp4) : ;;
-    *)       # default to mp3 if no/unknown extension (keeps your original intent)
+    *) ext="mp3" ;;   # default if unknown
-             ext="mp3"
-             ;;
   esac
   SOURCE_PATH="$WORKDIR/input.$ext"
-  echo "Downloading: $INPUT"
+  echo "↓ Downloading: $INPUT"
   wget -q --show-progress -O "$SOURCE_PATH" "$INPUT"
 else
-  # Local file
+  [[ -f "$INPUT" ]] || { echo "Error: file not found: $INPUT" >&2; exit 1; }
-  if [[ ! -f "$INPUT" ]]; then
-    echo "Error: file not found: $INPUT" >&2
-    exit 1
-  fi
   SOURCE_PATH="$INPUT"
   ext="$(lower_ext "$SOURCE_PATH")"
 fi
 
-# --- transcode if needed ---
+# -------- Transcode if needed --------
 case "$ext" in
   mp3)
     MP3_PATH="$SOURCE_PATH"
     ;;
   mp4)
     MP3_PATH="$WORKDIR/audio.mp3"
-    echo "Transcoding MP4 → MP3 with ffmpeg..."
+    echo "🎞  Transcoding MP4 → MP3 with ffmpeg..."
-    # -vn: drop video; -q:a 2 ~ VBR high quality (use -b:a 128k if you prefer CBR)
     ffmpeg -y -i "$SOURCE_PATH" -vn -acodec libmp3lame -q:a 2 "$MP3_PATH" >/dev/null 2>&1
     ;;
   *)
@@ -79,6 +117,12 @@ case "$ext" in
     ;;
 esac
 
-# --- run whisper ---
+# -------- Run whisper --------
-echo "Running whisper on: $MP3_PATH"
+OUT_FLAGS="$(output_flags_for "$OUTPUT_FORMAT")"
-"$WHISPER_EXEC" "$MP3_PATH" --model "$MODEL" --device cuda --language "$LANGUAGE"
+echo "▶ Running whisper"
+echo "   model:    $MODEL"
+echo "   language: $LANGUAGE"
+echo "   outputs:  $OUTPUT_FORMAT"
+echo "   input:    $MP3_PATH"
+# shellcheck disable=SC2086
+"$WHISPER_EXEC" "$MP3_PATH" --model "$MODEL" --device cuda --language "$LANGUAGE" $OUT_FLAGS