Files
personal_development/video_transcription/ai_transcriber/main.py
T
2026-01-10 17:08:51 -05:00

469 lines
23 KiB
Python

import argparse
import os
import sys
from dotenv import load_dotenv
# Load environment variables from central .env_files directory
# Path: .../personal_development/video_transcription/ai_transcriber/main.py
# Target: .../personal_development/.env_files/.env.aitranscribe
script_dir = os.path.dirname(os.path.abspath(__file__))
env_path = os.path.abspath(os.path.join(script_dir, '../../.env_files/.env.aitranscribe'))
if os.path.exists(env_path):
load_dotenv(env_path)
# print(f"Loaded configuration from: {env_path}") # Optional: Uncomment for debugging
else:
# Fallback: check local .env
local_env = os.path.join(script_dir, '.env')
if os.path.exists(local_env):
load_dotenv(local_env)
else:
# Last resort: just try loading generic (cwd)
load_dotenv()
from extractor import extract_audio, embed_subtitles
from transcriber import transcribe_audio, save_as_srt
from translator import translate_srt
from utils import validate_and_repair_srt
from diarizer import diarize_audio, merge_diarization_with_transcript
def save_srt_with_speakers(segments, output_path):
"""Helper to save SRT with speaker labels prepended to text."""
def format_timestamp(seconds: float):
whole_seconds = int(seconds)
milliseconds = int((seconds - whole_seconds) * 1000)
hours = whole_seconds // 3600
minutes = (whole_seconds % 3600) // 60
seconds = whole_seconds % 60
return f"{hours:02d}:{minutes:02d}:{seconds:02d},{milliseconds:03d}"
with open(output_path, "w", encoding="utf-8") as f:
for i, segment in enumerate(segments, start=1):
start = format_timestamp(segment["start"])
end = format_timestamp(segment["end"])
text = segment["text"].strip()
speaker = segment.get("speaker", "")
# Prepend speaker if present and not "Unknown"
if speaker and speaker != "Unknown":
text = f"[{speaker}]: {text}"
f.write(f"{i}\n")
f.write(f"{start} --> {end}\n")
f.write(f"{text}\n\n")
print(f"SRT saved to: {output_path}")
def process_file(file_path, args, source_lang=None):
print(f"\n=== Processing: {file_path} ===")
# 1. Extract Audio
audio_path = extract_audio(file_path)
# 2. Transcribe (Generate SRT)
transcript_file = os.path.splitext(file_path)[0] + ".srt"
transcript_exists = os.path.exists(transcript_file) and not args.force
# Variable to hold final SRT path for embedding
final_srt_path = transcript_file
if transcript_exists:
print(f"Transcript exists: {transcript_file}. Skipping transcription.")
with open(transcript_file, "r", encoding="utf-8") as f:
srt_content = f.read()
else:
# Transcribe
result = transcribe_audio(audio_path, model_size=args.model, language=source_lang)
segments = result["segments"]
# Optional: Diarization
if args.diarize:
hf_token = args.hf_token or os.getenv("HF_TOKEN")
if hf_token:
print("Running Speaker Diarization...")
diar_segments = diarize_audio(audio_path, hf_token=hf_token)
if diar_segments:
segments = merge_diarization_with_transcript(segments, diar_segments)
print("Diarization merged into transcript.")
else:
print("Warning: --diarize requested but HF_TOKEN not provided. Skipping.")
# Save SRT
# Use simple save if no speakers, or custom if speakers
if args.diarize:
save_srt_with_speakers(segments, transcript_file)
else:
save_as_srt(result, transcript_file)
# Validation
validate_and_repair_srt(transcript_file)
with open(transcript_file, "r", encoding="utf-8") as f:
srt_content = f.read()
# 3. Translate (Generate Translated SRT)
translated_file = os.path.splitext(file_path)[0] + f".{args.lang}.srt"
translation_success = False
if os.path.exists(translated_file) and not args.force:
print(f"Translation exists: {translated_file}. Skipping translation.")
final_srt_path = translated_file
translation_success = True
else:
# Only translate if there is content
if srt_content:
translated_srt_content = translate_srt(srt_content, target_language=args.lang)
if translated_srt_content:
with open(translated_file, "w", encoding="utf-8") as f:
f.write(translated_srt_content)
print(f"Translation saved to: {translated_file}")
validate_and_repair_srt(translated_file)
final_srt_path = translated_file
translation_success = True
else:
print("⚠️ TRANSLATION FAILED.")
translation_success = False
# 4. Embed Subtitles
# SAFETY: If translation was intended but failed, do NOT embed/delete to prevent
# replacing the video with one containing only untranslated subtitles.
should_embed = args.embed
if args.embed and not translation_success:
print("\n❌ SAFETY HALT: Translation failed. Skipping embedding and deletion to preserve original file.")
should_embed = False
if should_embed:
embed_subtitles(file_path, final_srt_path)
import argparse
import os
import sys
from dotenv import load_dotenv
# Load environment variables from central .env_files directory
# Path: .../personal_development/video_transcription/ai_transcriber/main.py
# Target: .../personal_development/.env_files/.env.aitranscribe
script_dir = os.path.dirname(os.path.abspath(__file__))
env_path = os.path.abspath(os.path.join(script_dir, '../../.env_files/.env.aitranscribe'))
if os.path.exists(env_path):
load_dotenv(env_path)
print(f"Loaded configuration from: {env_path}")
else:
# Fallback: check local .env
local_env = os.path.join(script_dir, '.env')
if os.path.exists(local_env):
load_dotenv(local_env)
else:
# Last resort: just try loading generic (cwd)
load_dotenv()
from extractor import extract_audio, embed_subtitles
from transcriber import transcribe_audio, save_as_srt
from translator import translate_srt
from utils import validate_and_repair_srt
from diarizer import diarize_audio, merge_diarization_with_transcript
import tracker
from tracker import JobStatus
def save_srt_with_speakers(segments, output_path):
"""Helper to save SRT with speaker labels prepended to text."""
def format_timestamp(seconds: float):
whole_seconds = int(seconds)
milliseconds = int((seconds - whole_seconds) * 1000)
hours = whole_seconds // 3600
minutes = (whole_seconds % 3600) // 60
seconds = whole_seconds % 60
return f"{hours:02d}:{minutes:02d}:{seconds:02d},{milliseconds:03d}"
with open(output_path, "w", encoding="utf-8") as f:
for i, segment in enumerate(segments, start=1):
start = format_timestamp(segment["start"])
end = format_timestamp(segment["end"])
text = segment["text"].strip()
speaker = segment.get("speaker", "")
# Prepend speaker if present and not "Unknown"
if speaker and speaker != "Unknown":
text = f"[{speaker}]: {text}"
f.write(f"{i}\n")
f.write(f"{start} --> {end}\n")
f.write(f"{text}\n\n")
print(f"SRT saved to: {output_path}")
def process_file(file_path, args, source_lang=None):
tracker.logger.info(f"=== Processing: {file_path} ===")
# Initialize Job
job = tracker.get_job(file_path)
if job.status == JobStatus.COMPLETED and not args.force:
tracker.logger.info("Job already completed. Skipping.")
return
tracker.update_job_status(file_path, JobStatus.PROCESSING)
try:
# 1. Extract Audio
tracker.update_step(file_path, "step_extract", "processing")
audio_path = extract_audio(file_path)
tracker.update_step(file_path, "step_extract", "done")
# 2. Transcribe (Generate SRT)
tracker.update_step(file_path, "step_transcribe", "processing")
transcript_file = os.path.splitext(file_path)[0] + ".srt"
transcript_exists = os.path.exists(transcript_file) and not args.force
# Variable to hold final SRT path for embedding
final_srt_path = transcript_file
if transcript_exists:
tracker.logger.info(f"Transcript exists: {transcript_file}. Skipping transcription.")
with open(transcript_file, "r", encoding="utf-8") as f:
srt_content = f.read()
else:
# Transcribe
result = transcribe_audio(audio_path, model_size=args.model, language=source_lang)
segments = result["segments"]
# Optional: Diarization
if args.diarize:
hf_token = args.hf_token or os.getenv("HF_TOKEN")
if hf_token:
tracker.logger.info("Running Speaker Diarization...")
diar_segments = diarize_audio(audio_path, hf_token=hf_token)
if diar_segments:
segments = merge_diarization_with_transcript(segments, diar_segments)
tracker.logger.info("Diarization merged into transcript.")
else:
tracker.logger.warning("Warning: --diarize requested but HF_TOKEN not provided. Skipping.")
# Save SRT
if args.diarize:
save_srt_with_speakers(segments, transcript_file)
else:
save_as_srt(result, transcript_file)
# Validation
validate_and_repair_srt(transcript_file)
with open(transcript_file, "r", encoding="utf-8") as f:
srt_content = f.read()
tracker.update_step(file_path, "step_transcribe", "done")
# 3. Translate (Generate Translated SRT)
tracker.update_step(file_path, "step_translate", "processing")
translated_file = os.path.splitext(file_path)[0] + f".{args.lang}.srt"
translation_success = False
if os.path.exists(translated_file) and not args.force:
tracker.logger.info(f"Translation exists: {translated_file}. Skipping translation.")
final_srt_path = translated_file
translation_success = True
else:
# Only translate if there is content
if srt_content:
translated_srt_content = translate_srt(srt_content, target_language=args.lang)
if translated_srt_content:
with open(translated_file, "w", encoding="utf-8") as f:
f.write(translated_srt_content)
tracker.logger.info(f"Translation saved to: {translated_file}")
validate_and_repair_srt(translated_file)
final_srt_path = translated_file
translation_success = True
else:
tracker.logger.error("TRANSLATION FAILED.")
tracker.update_step(file_path, "step_translate", "failed")
translation_success = False
if translation_success:
tracker.update_step(file_path, "step_translate", "done")
# 4. Embed Subtitles
tracker.update_step(file_path, "step_embed", "processing")
should_embed = args.embed
if args.embed and not translation_success:
tracker.logger.warning("SAFETY HALT: Translation failed. Skipping embedding and deletion to preserve original file.")
should_embed = False
if should_embed:
embed_subtitles(file_path, final_srt_path)
# 5. Delete Source File (Optional & Risky)
if args.delete_source:
if args.embed:
# Safety: Ensure the new subbed video exists before deleting the old one
base, ext = os.path.splitext(file_path)
expected_output = f"{base}.subbed{ext}"
if os.path.exists(expected_output):
try:
os.remove(file_path)
tracker.logger.info(f"SOURCE DELETED: Original file '{file_path}' has been removed.")
except OSError as e:
tracker.logger.error(f"Error: Could not delete source file: {e}")
else:
tracker.logger.error(f"SAFETY ABORT: Source file NOT deleted. Could not find expected output '{expected_output}'.")
else:
tracker.logger.warning("SAFETY ABORT: Source file NOT deleted. You must enable --embed to safely replace the video.")
tracker.update_step(file_path, "step_embed", "done")
# 5. Cleanup Audio
if args.cleanup:
try:
os.remove(audio_path)
tracker.logger.info(f"Cleanup: Removed temporary audio file {audio_path}")
except OSError as e:
tracker.logger.warning(f"Warning: Could not remove audio file: {e}")
# Mark Complete
if translation_success:
tracker.update_job_status(file_path, JobStatus.COMPLETED)
else:
# If translation failed but we didn't crash, we technically finished the run but result is partial
tracker.update_job_status(file_path, JobStatus.FAILED, error="Translation failed")
except Exception as e:
tracker.logger.exception(f"Job Failed for {file_path}")
tracker.update_job_status(file_path, JobStatus.FAILED, error=str(e))
# Don't exit, allow other files to process
return
def main():
parser = argparse.ArgumentParser(description="AI Video Transcriber & Translator")
parser.add_argument("input", nargs='?', help="Path to video file or directory")
parser.add_argument("--model", default="auto", choices=["auto", "tiny", "base", "small", "medium", "large"], help="Whisper model size (default: auto)")
parser.add_argument("--lang", default="English", help="Target language for translation (default: English)")
parser.add_argument("--source-lang", help="Source language of the audio (e.g., 'fr', 'es'). If omitted, you will be prompted.")
parser.add_argument("--force", action="store_true", help="Overwrite existing transcript/translation files")
# New Arguments
parser.add_argument("--cleanup", action="store_true", help="Delete the temporary .wav file after processing")
parser.add_argument("--embed", action="store_true", help="Embed the final subtitles into the video (Soft Subs)")
parser.add_argument("--diarize", action="store_true", help="Enable speaker diarization (requires HF_TOKEN)")
parser.add_argument("--hf-token", help="HuggingFace Token for pyannote.audio (or set HF_TOKEN env var)")
parser.add_argument("--delete-source", action="store_true", help="Delete the original video file AFTER successful embedding")
parser.add_argument("--retry-failed", action="store_true", help="Retry only jobs marked as FAILED in the database")
args = parser.parse_args()
if not os.getenv("GEMINI_API_KEY"):
print("Warning: GEMINI_API_KEY environment variable not set. Translation step will fail.")
# Handling Retry Logic
if args.retry_failed:
print("Retrying failed jobs from database...")
failed_files = tracker.get_failed_jobs()
if not failed_files:
print("No failed jobs found.")
return
# We need args.source_lang logic here too if needed, but for retries we might assume context
# For simplicity, we'll prompt if missing just like normal run
# Determine source language (Prompt if missing)
source_lang = args.source_lang
if not source_lang:
print("\n--- Audio Configuration ---")
user_input = input("Enter the source language of the video(s) (e.g., 'French', 'es').\nPress Enter to use Whisper's auto-detection: ").strip()
if user_input:
source_lang = user_input
else:
source_lang = None # Let Whisper auto-detect
print("Selected: Auto-detect")
for file_path in failed_files:
if os.path.exists(file_path):
process_file(file_path, args, source_lang)
else:
print(f"Skipping missing file: {file_path}")
return
# Normal Logic
if not args.input:
parser.print_help()
sys.exit(1)
# Determine source language (Prompt if missing)
source_lang = args.source_lang
if not source_lang:
print("\n--- Audio Configuration ---")
user_input = input("Enter the source language of the video(s) (e.g., 'French', 'es').\nPress Enter to use Whisper's auto-detection: ").strip()
if user_input:
source_lang = user_input
else:
source_lang = None # Let Whisper auto-detect
print("Selected: Auto-detect")
if os.path.isfile(args.input):
process_file(args.input, args, source_lang)
elif os.path.isdir(args.input):
video_extensions = ('.mp4', '.mkv', '.mov', '.avi', '.webm', '.flv', '.wmv', '.m4v')
found_files = False
for root, dirs, files in os.walk(args.input):
for file in files:
if file.lower().endswith(video_extensions):
found_files = True
file_path = os.path.join(root, file)
process_file(file_path, args, source_lang)
if not found_files:
print(f"No video files found in {args.input}")
else:
print(f"Error: Invalid input path '{args.input}'")
sys.exit(1)
if __name__ == "__main__":
main()
def main():
parser = argparse.ArgumentParser(description="AI Video Transcriber & Translator")
parser.add_argument("input", help="Path to video file or directory")
parser.add_argument("--model", default="auto", choices=["auto", "tiny", "base", "small", "medium", "large"], help="Whisper model size (default: auto)")
parser.add_argument("--lang", default="English", help="Target language for translation (default: English)")
parser.add_argument("--source-lang", help="Source language of the audio (e.g., 'fr', 'es'). If omitted, you will be prompted.")
parser.add_argument("--force", action="store_true", help="Overwrite existing transcript/translation files")
# New Arguments
parser.add_argument("--cleanup", action="store_true", help="Delete the temporary .wav file after processing")
parser.add_argument("--embed", action="store_true", help="Embed the final subtitles into the video (Soft Subs)")
parser.add_argument("--diarize", action="store_true", help="Enable speaker diarization (requires HF_TOKEN)")
parser.add_argument("--hf-token", help="HuggingFace Token for pyannote.audio (or set HF_TOKEN env var)")
parser.add_argument("--delete-source", action="store_true", help="Delete the original video file AFTER successful embedding")
args = parser.parse_args()
if not os.getenv("GEMINI_API_KEY"):
print("Warning: GEMINI_API_KEY environment variable not set. Translation step will fail.")
# Determine source language (Prompt if missing)
source_lang = args.source_lang
if not source_lang:
print("\n--- Audio Configuration ---")
user_input = input("Enter the source language of the video(s) (e.g., 'French', 'es').\nPress Enter to use Whisper's auto-detection: ").strip()
if user_input:
source_lang = user_input
else:
source_lang = None # Let Whisper auto-detect
print("Selected: Auto-detect")
if os.path.isfile(args.input):
process_file(args.input, args, source_lang)
elif os.path.isdir(args.input):
video_extensions = ('.mp4', '.mkv', '.mov', '.avi', '.webm', '.flv', '.wmv', '.m4v')
found_files = False
for root, dirs, files in os.walk(args.input):
for file in files:
if file.lower().endswith(video_extensions):
found_files = True
file_path = os.path.join(root, file)
process_file(file_path, args, source_lang)
if not found_files:
print(f"No video files found in {args.input}")
else:
print(f"Error: Invalid input path '{args.input}'")
sys.exit(1)
if __name__ == "__main__":
main()