Jamie
7e373cc6c4
All checks were successful
Build and Push / build (push) Successful in 11m39s
Fixing ASCII encoding error by forcing UTF-8
61 lines
2.1 KiB
Python
61 lines
2.1 KiB
Python
import json
|
|
from pathlib import Path
|
|
|
|
|
|
def fmt_time(seconds: float) -> str:
|
|
h, rem = divmod(int(seconds), 3600)
|
|
m, s = divmod(rem, 60)
|
|
return f"{h:02d}:{m:02d}:{s:02d}"
|
|
|
|
|
|
def load_turns(transcript_json: Path, speaker_map: dict[str, str] | None = None) -> list[dict]:
|
|
"""Parse whisperx json into merged (start, end, speaker, text) turns.
|
|
speaker_map maps raw labels (SPEAKER_00...) to display names; unmapped labels pass through raw."""
|
|
data = json.loads(transcript_json.read_text(encoding="utf-8"))
|
|
speaker_map = speaker_map or {}
|
|
turns = []
|
|
for seg in data.get("segments", []):
|
|
raw = seg.get("speaker", "UNKNOWN")
|
|
speaker = speaker_map.get(raw, raw)
|
|
text = (seg.get("text") or "").strip()
|
|
if not text:
|
|
continue
|
|
if turns and turns[-1]["speaker"] == speaker:
|
|
turns[-1]["text"] += " " + text
|
|
turns[-1]["end"] = seg["end"]
|
|
else:
|
|
turns.append({"start": seg["start"], "end": seg["end"], "speaker": speaker, "text": text, "raw_speaker": raw})
|
|
return turns
|
|
|
|
|
|
def distinct_speakers(transcript_json: Path) -> list[dict]:
|
|
"""Return each raw speaker label with a couple of sample lines, for the naming UI."""
|
|
turns = load_turns(transcript_json) # no map -> raw labels
|
|
seen: dict[str, list[dict]] = {}
|
|
for t in turns:
|
|
seen.setdefault(t["raw_speaker"], []).append(t)
|
|
return [
|
|
{
|
|
"raw_label": label,
|
|
"samples": [{"start": t["start"], "text": t["text"][:140]} for t in items[:3]],
|
|
}
|
|
for label, items in sorted(seen.items())
|
|
]
|
|
|
|
|
|
def chunk_turns(turns: list[dict], word_target: int) -> list[list[dict]]:
|
|
chunks, current, word_count = [], [], 0
|
|
for turn in turns:
|
|
current.append(turn)
|
|
word_count += len(turn["text"].split())
|
|
if word_count >= word_target:
|
|
chunks.append(current)
|
|
current, word_count = [], 0
|
|
if current:
|
|
chunks.append(current)
|
|
return chunks
|
|
|
|
|
|
def turns_to_text(turns: list[dict]) -> str:
|
|
return "\n".join(f"[{fmt_time(t['start'])}] {t['speaker']}: {t['text']}" for t in turns)
|