fix(video2srt): format to two lines
This commit is contained in:
@@ -9,6 +9,8 @@ TRANSLATE="${TRANSLATE:-1}"
|
||||
SRC_LANG="${SRC_LANG:-auto}"
|
||||
FORCE="${FORCE:-0}"
|
||||
USE_VAD="${USE_VAD:-1}"
|
||||
MAX_LEN="${MAX_LEN:-84}"
|
||||
LINE_LEN="${LINE_LEN:-42}"
|
||||
OUTPUT=""
|
||||
|
||||
usage() {
|
||||
@@ -24,6 +26,8 @@ Options:
|
||||
-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)
|
||||
--max-len N Max characters per SRT entry, 0 to disable (default: $MAX_LEN)
|
||||
--line-len N Max characters per visible line; longer entries wrap to 2 lines, 0 to disable (default: $LINE_LEN)
|
||||
-h, --help Show this help
|
||||
|
||||
Env overrides: WHISPER_DIR, WHISPER_MODEL, WHISPER_BIN
|
||||
@@ -39,6 +43,8 @@ while [[ $# -gt 0 ]]; do
|
||||
-o|--output) OUTPUT="$2"; shift 2 ;;
|
||||
-f|--force) FORCE=1; shift ;;
|
||||
--no-vad) USE_VAD=0; shift ;;
|
||||
--max-len) MAX_LEN="$2"; shift 2 ;;
|
||||
--line-len) LINE_LEN="$2"; shift 2 ;;
|
||||
-h|--help) usage; exit 0 ;;
|
||||
--) shift; args+=("$@"); break ;;
|
||||
-*) echo "Unknown option: $1" >&2; usage >&2; exit 2 ;;
|
||||
@@ -96,6 +102,9 @@ for video in "${args[@]}"; do
|
||||
-mc 0
|
||||
)
|
||||
[[ "$TRANSLATE" == "1" ]] && whisper_args+=(--translate)
|
||||
if [[ "$MAX_LEN" -gt 0 ]]; then
|
||||
whisper_args+=(--max-len "$MAX_LEN" --split-on-word)
|
||||
fi
|
||||
if [[ "$USE_VAD" == "1" ]]; then
|
||||
if [[ -f "$VAD_MODEL" ]]; then
|
||||
whisper_args+=(--vad --vad-model "$VAD_MODEL")
|
||||
@@ -107,6 +116,50 @@ for video in "${args[@]}"; do
|
||||
|
||||
"$WHISPER_BIN" "${whisper_args[@]}"
|
||||
|
||||
if [[ "$LINE_LEN" -gt 0 ]]; then
|
||||
python3 - "$srt" "$LINE_LEN" <<'PYEOF'
|
||||
import re, sys, pathlib
|
||||
|
||||
path = pathlib.Path(sys.argv[1])
|
||||
max_line = int(sys.argv[2])
|
||||
|
||||
def wrap(text, limit):
|
||||
text = " ".join(text.split())
|
||||
if len(text) <= limit:
|
||||
return text
|
||||
words = text.split(" ")
|
||||
if len(words) < 2:
|
||||
return text
|
||||
best = None
|
||||
best_score = None
|
||||
cum = 0
|
||||
for i, w in enumerate(words[:-1]):
|
||||
cum += len(w) + (1 if i > 0 else 0)
|
||||
top, bot = cum, len(text) - cum - 1
|
||||
score = abs(bot - top)
|
||||
if top > limit:
|
||||
score += (top - limit) * 100
|
||||
if bot > limit:
|
||||
score += (bot - limit) * 100
|
||||
if w.rstrip(",.!?:;") != w:
|
||||
score -= 8
|
||||
if best_score is None or score < best_score:
|
||||
best_score, best = score, i
|
||||
return " ".join(words[:best+1]) + "\n" + " ".join(words[best+1:])
|
||||
|
||||
blocks = re.split(r"\n\n+", path.read_text(encoding="utf-8").strip())
|
||||
out = []
|
||||
for b in blocks:
|
||||
lines = b.split("\n")
|
||||
if len(lines) < 3:
|
||||
out.append(b)
|
||||
continue
|
||||
head, body = lines[:2], " ".join(lines[2:])
|
||||
out.append("\n".join(head + [wrap(body, max_line)]))
|
||||
path.write_text("\n\n".join(out) + "\n", encoding="utf-8")
|
||||
PYEOF
|
||||
fi
|
||||
|
||||
rm -f "$tmpwav"
|
||||
trap - EXIT
|
||||
echo "<< $srt"
|
||||
|
||||
Reference in New Issue
Block a user