Skip to content

Open-source AI podcast clipper. Generate vertical clips with face tracking and burned-in captions. CLI, MCP server, web app.

Let studio take a transcript instead of running Whisper (#209)

Nika Siradze1f5e48c96255
diff --git a/backend/clip_studio.py b/backend/clip_studio.py
index 8a12dd9..8e79196 100644
--- a/backend/clip_studio.py
+++ b/backend/clip_studio.py
@@ -96,6 +96,37 @@ def _transcribe(video: str, language: str | None, engine: str | None):
return res.get("words", [])
+def _load_transcript(path: str) -> list:
+ import json
+ import math
+ try:
+ with open(path, encoding="utf-8") as fh:
+ data = json.load(fh)
+ except (OSError, UnicodeDecodeError, json.JSONDecodeError) as exc:
+ raise SystemExit(f"--transcript {path}: {exc}")
+ if isinstance(data, dict):
+ if "words" not in data:
+ raise SystemExit(f"--transcript {path}: expected {{\"words\": [...]}}")
+ words = data["words"]
+ else:
+ words = data
+ if not isinstance(words, list):
+ raise SystemExit(f"--transcript {path}: expected a list of words or {{\"words\": [...]}}")
+ out = []
+ for w in words:
+ try:
+ start, end = float(w["start"]), float(w["end"])
+ except (KeyError, TypeError, ValueError):
+ continue
+ if not (math.isfinite(start) and math.isfinite(end)) or start < 0 or end < start:
+ continue
+ out.append({"word": str(w.get("word", "")), "start": start, "end": end})
+ if not out:
+ raise SystemExit(f"--transcript {path}: no usable words in it")
+ print(f" [transcript] using {len(out)} supplied words", flush=True)
+ return out
+
+
def _find_paragraph(words: list, phrase: str) -> tuple[float, float]:
"""Find the time span of a paragraph by fuzzy-matching its text against
the transcript word stream. Returns (start, end) seconds."""
@@ -234,6 +265,7 @@ def main():
ap.add_argument("--paragraph", help="Find fragment by matching this text in the transcript")
ap.add_argument("--language", default=None, help="Transcription language (e.g. es). Auto-detect if omitted.")
ap.add_argument("--engine", choices=["whisper-py", "whispercpp", "assemblyai"], default=None, help="Transcription engine")
+ ap.add_argument("--transcript", default=None, help="Word timings JSON for this video (list of {word,start,end} or {words:[...]}); skips transcription")
ap.add_argument("--caption-style", default="hormozi", choices=["hormozi", "karaoke", "subtle", "branded"])
ap.add_argument("--caption-position", default="auto", choices=["auto", "upper", "center", "lower"])
ap.add_argument("--caption-scale", type=float, default=1.0)
@@ -313,8 +345,7 @@ def main():
out_dir = os.path.join(data_dir, "output")
os.makedirs(out_dir, exist_ok=True)
- # Need a transcript if cutting by paragraph or if rendering captions.
- words = _transcribe(video, args.language, args.engine)
+ words = _load_transcript(args.transcript) if args.transcript else _transcribe(video, args.language, args.engine)
if args.paragraph:
start, end = _find_paragraph(words, args.paragraph)
diff --git a/backend/services/transcription.py b/backend/services/transcription.py
index 1a5d4af..e6859e9 100644
--- a/backend/services/transcription.py
+++ b/backend/services/transcription.py
@@ -58,6 +58,15 @@ def _whispercpp_ready(model_size: str) -> bool:
return _whispercpp_cli() is not None and os.path.exists(_whispercpp_model(model_size))
+def _whisper_threads() -> int:
+ raw = os.environ.get("PODCLI_WHISPER_THREADS", "").strip()
+ try:
+ n = int(raw)
+ except ValueError:
+ n = 0
+ return n if n > 0 else 4
+
+
def _transcribe_with_whispercpp(file_path, model_size, language, progress_callback, wav_path=None):
from services import transcription_whispercpp as wcpp
@@ -81,6 +90,7 @@ def _transcribe_with_whispercpp(file_path, model_size, language, progress_callba
vad=vad,
vad_model=os.environ.get("PODCLI_WHISPERCPP_VAD_MODEL") or None,
wav_path=wav_path,
+ threads=_whisper_threads(),
)
if progress_callback:
progress_callback(50, "Transcription complete")