From ffbdcc197058b58ff20ae33d0024fafe37aef500 Mon Sep 17 00:00:00 2001
From: Oscar Wallberg <oscar.wallberg@outlook.com>
Date: Mon, 25 May 2026 17:29:20 +0200
Subject: [PATCH] feat(scripts): add video2srt

---
 .local/bin/video2srt | 113 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 113 insertions(+)
 create mode 100755 .local/bin/video2srt

diff --git a/.local/bin/video2srt b/.local/bin/video2srt
new file mode 100755
index 0000000..ff6a3c1
--- /dev/null
+++ b/.local/bin/video2srt
@@ -0,0 +1,113 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+WHISPER_DIR="${WHISPER_DIR:-$HOME/repos/whisper.cpp}"
+WHISPER_MODEL="${WHISPER_MODEL:-$WHISPER_DIR/models/ggml-large-v3.bin}"
+WHISPER_BIN="${WHISPER_BIN:-$WHISPER_DIR/build/bin/whisper-cli}"
+VAD_MODEL="${VAD_MODEL:-$WHISPER_DIR/models/ggml-silero-v6.2.0.bin}"
+TRANSLATE="${TRANSLATE:-1}"
+SRC_LANG="${SRC_LANG:-auto}"
+FORCE="${FORCE:-0}"
+USE_VAD="${USE_VAD:-1}"
+OUTPUT=""
+
+usage() {
+    cat <<EOF
+Usage: video2srt [options] <video> [<video>...]
+
+Generates an English .srt next to each video using whisper.cpp.
+
+Options:
+  -t, --transcribe   Transcribe in source language (default: translate to English)
+  -l, --lang CODE    Force source language (default: auto-detect)
+  -m, --model PATH   Path to ggml model (default: $WHISPER_MODEL)
+  -o, --output PATH  Output .srt path (single input only; default: <video>.srt)
+  -f, --force        Overwrite existing .srt
+      --no-vad       Disable Silero VAD pre-filtering (VAD reduces hallucination loops)
+  -h, --help         Show this help
+
+Env overrides: WHISPER_DIR, WHISPER_MODEL, WHISPER_BIN
+EOF
+}
+
+args=()
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        -t|--transcribe) TRANSLATE=0; shift ;;
+        -l|--lang) SRC_LANG="$2"; shift 2 ;;
+        -m|--model) WHISPER_MODEL="$2"; shift 2 ;;
+        -o|--output) OUTPUT="$2"; shift 2 ;;
+        -f|--force) FORCE=1; shift ;;
+        --no-vad) USE_VAD=0; shift ;;
+        -h|--help) usage; exit 0 ;;
+        --) shift; args+=("$@"); break ;;
+        -*) echo "Unknown option: $1" >&2; usage >&2; exit 2 ;;
+        *) args+=("$1"); shift ;;
+    esac
+done
+
+if [[ ${#args[@]} -eq 0 ]]; then
+    usage >&2; exit 2
+fi
+
+if [[ -n "$OUTPUT" && ${#args[@]} -gt 1 ]]; then
+    echo "--output cannot be combined with multiple input files" >&2; exit 2
+fi
+
+[[ -x "$WHISPER_BIN" ]] || { echo "whisper-cli not found at $WHISPER_BIN" >&2; exit 1; }
+[[ -f "$WHISPER_MODEL" ]] || { echo "model not found at $WHISPER_MODEL" >&2; exit 1; }
+command -v ffmpeg >/dev/null || { echo "ffmpeg not installed" >&2; exit 1; }
+
+for video in "${args[@]}"; do
+    if [[ ! -f "$video" ]]; then
+        echo "skip: $video (not a file)" >&2
+        continue
+    fi
+
+    if [[ -n "$OUTPUT" ]]; then
+        out_stem="${OUTPUT%.srt}"
+        out_dir=$(dirname -- "$out_stem")
+    else
+        out_dir=$(dirname -- "$video")
+        base=$(basename -- "$video")
+        out_stem="$out_dir/${base%.*}"
+    fi
+    srt="$out_stem.srt"
+
+    if [[ -f "$srt" && "$FORCE" != "1" ]]; then
+        echo "skip: $srt exists (use --force to overwrite)"
+        continue
+    fi
+
+    echo ">> $video"
+    tmpwav=$(mktemp --suffix=.wav)
+    trap 'rm -f "$tmpwav"' EXIT
+
+    ffmpeg -hide_banner -loglevel error -y \
+        -i "$video" -vn -ar 16000 -ac 1 -c:a pcm_s16le "$tmpwav"
+
+    mkdir -p -- "$out_dir"
+    whisper_args=(
+        -m "$WHISPER_MODEL"
+        -f "$tmpwav"
+        -of "$out_stem"
+        --output-srt
+        -l "$SRC_LANG"
+        -mc 0
+    )
+    [[ "$TRANSLATE" == "1" ]] && whisper_args+=(--translate)
+    if [[ "$USE_VAD" == "1" ]]; then
+        if [[ -f "$VAD_MODEL" ]]; then
+            whisper_args+=(--vad --vad-model "$VAD_MODEL")
+        else
+            echo "warn: VAD model not found at $VAD_MODEL, running without VAD" >&2
+            echo "      download with: sh $WHISPER_DIR/models/download-vad-model.sh silero-v6.2.0" >&2
+        fi
+    fi
+
+    "$WHISPER_BIN" "${whisper_args[@]}"
+
+    rm -f "$tmpwav"
+    trap - EXIT
+    echo "<< $srt"
+done