Skip to content

Speakr is a personal, self-hosted web application designed for transcribing audio recordings

Python

Add automatic audio compression for lossless uploads (#152)

Murtaza Nasire6b3b7ef934c
diff --git a/config/env.asr.example b/config/env.asr.example
index 1d1845a7..e9195b25 100644
--- a/config/env.asr.example
+++ b/config/env.asr.example
@@ -85,6 +85,16 @@ TIMEZONE="UTC"
# Options: DEBUG, INFO, WARNING, ERROR
LOG_LEVEL="INFO"
+# --- Audio Compression ---
+# Automatically compress lossless uploads (WAV, AIFF) to save storage
+AUDIO_COMPRESS_UPLOADS=true
+
+# Target codec: mp3 (lossy, smallest), flac (lossless), opus (lossy, efficient)
+AUDIO_CODEC=mp3
+
+# Bitrate for lossy codecs (ignored for FLAC)
+AUDIO_BITRATE=128k
+
# --- Admin User (created on first run) ---
ADMIN_USERNAME=admin
diff --git a/config/env.whisper.example b/config/env.whisper.example
index 3d357388..862b30f4 100644
--- a/config/env.whisper.example
+++ b/config/env.whisper.example
@@ -87,6 +87,16 @@ CHUNK_LIMIT=20MB
# Overlap between chunks (seconds)
CHUNK_OVERLAP_SECONDS=3
+# --- Audio Compression ---
+# Automatically compress lossless uploads (WAV, AIFF) to save storage
+AUDIO_COMPRESS_UPLOADS=true
+
+# Target codec: mp3 (lossy, smallest), flac (lossless), opus (lossy, efficient)
+AUDIO_CODEC=mp3
+
+# Bitrate for lossy codecs (ignored for FLAC)
+AUDIO_BITRATE=128k
+
# --- Admin User (created on first run) ---
ADMIN_USERNAME=admin
diff --git a/config/env.whisperx.example b/config/env.whisperx.example
index f6e2eb46..c524b707 100644
--- a/config/env.whisperx.example
+++ b/config/env.whisperx.example
@@ -97,6 +97,16 @@ TIMEZONE="UTC"
# Options: DEBUG, INFO, WARNING, ERROR
LOG_LEVEL="INFO"
+# --- Audio Compression ---
+# Automatically compress lossless uploads (WAV, AIFF) to save storage
+AUDIO_COMPRESS_UPLOADS=true
+
+# Target codec: mp3 (lossy, smallest), flac (lossless), opus (lossy, efficient)
+AUDIO_CODEC=mp3
+
+# Bitrate for lossy codecs (ignored for FLAC)
+AUDIO_BITRATE=128k
+
# --- Admin User (created on first run) ---
ADMIN_USERNAME=admin
diff --git a/src/api/recordings.py b/src/api/recordings.py
index f7a3fbde..5c07ea02 100644
--- a/src/api/recordings.py
+++ b/src/api/recordings.py
@@ -22,8 +22,8 @@ from email.utils import encode_rfc2231
from src.database import db
from src.models import *
from src.utils import *
-from src.config.app_config import ASR_MIN_SPEAKERS, ASR_MAX_SPEAKERS, ASR_DIARIZE
-from src.tasks.processing import format_transcription_for_llm, transcribe_with_chunking
+from src.config.app_config import ASR_MIN_SPEAKERS, ASR_MAX_SPEAKERS, ASR_DIARIZE, AUDIO_COMPRESS_UPLOADS, AUDIO_CODEC, AUDIO_BITRATE
+from src.tasks.processing import format_transcription_for_llm, transcribe_with_chunking, compress_lossless_audio
from src.services.speaker import update_speaker_usage, identify_unidentified_speakers_from_text
from src.services.speaker_embedding_matcher import update_speaker_embedding
from src.services.speaker_snippets import create_speaker_snippets
@@ -2034,6 +2034,17 @@ def upload_file():
current_app.logger.error(f"ffmpeg conversion failed for {filepath}: {e.stderr}")
return jsonify({'error': f'Failed to convert audio file: {e.stderr}'}), 500
+ # Compress lossless audio files (WAV, AIFF) if enabled
+ if AUDIO_COMPRESS_UPLOADS:
+ try:
+ new_filepath, new_mime = compress_lossless_audio(filepath, AUDIO_CODEC, AUDIO_BITRATE)
+ if new_filepath != filepath:
+ filepath = new_filepath
+ # MIME type will be updated below
+ except Exception as e:
+ current_app.logger.warning(f"Audio compression failed, continuing with original: {e}")
+ # Don't fail the upload if compression fails - just continue with original
+
# Get final file size (of original or converted file)
final_file_size = os.path.getsize(filepath)
diff --git a/src/config/app_config.py b/src/config/app_config.py
index a078c8ca..e86376d7 100644
--- a/src/config/app_config.py
+++ b/src/config/app_config.py
@@ -43,6 +43,11 @@ ENABLE_CHUNKING = os.environ.get('ENABLE_CHUNKING', 'true').lower() == 'true'
CHUNK_SIZE_MB = int(os.environ.get('CHUNK_SIZE_MB', '20'))
CHUNK_OVERLAP_SECONDS = int(os.environ.get('CHUNK_OVERLAP_SECONDS', '3'))
+# Audio compression settings - compress lossless uploads (WAV, AIFF) to save storage
+AUDIO_COMPRESS_UPLOADS = os.environ.get('AUDIO_COMPRESS_UPLOADS', 'true').lower() == 'true'
+AUDIO_CODEC = os.environ.get('AUDIO_CODEC', 'mp3').lower() # mp3, flac, opus
+AUDIO_BITRATE = os.environ.get('AUDIO_BITRATE', '128k') # For lossy codecs
+
# Create chunking service at module level so it can be imported by processing.py
chunking_service = AudioChunkingService(CHUNK_SIZE_MB, CHUNK_OVERLAP_SECONDS) if ENABLE_CHUNKING else None
diff --git a/src/tasks/__init__.py b/src/tasks/__init__.py
index 4e52fbbe..000e5686 100644
--- a/src/tasks/__init__.py
+++ b/src/tasks/__init__.py
@@ -7,6 +7,7 @@ from .processing import (
generate_summary_only_task,
extract_events_from_transcript,
extract_audio_from_video,
+ compress_lossless_audio,
transcribe_audio_asr,
transcribe_audio_task,
transcribe_single_file,
@@ -18,6 +19,7 @@ __all__ = [
'generate_summary_only_task',
'extract_events_from_transcript',
'extract_audio_from_video',
+ 'compress_lossless_audio',
'transcribe_audio_asr',
'transcribe_audio_task',
'transcribe_single_file',
diff --git a/src/tasks/processing.py b/src/tasks/processing.py
index 9601449f..2a945126 100644
--- a/src/tasks/processing.py
+++ b/src/tasks/processing.py
@@ -874,6 +874,115 @@ def extract_audio_from_video(video_filepath, output_format='mp3', cleanup_origin
raise
+def compress_lossless_audio(filepath, codec='mp3', bitrate='128k'):
+ """Compress lossless audio files (WAV, AIFF) to save storage.
+
+ Only compresses lossless formats - already-compressed formats are skipped
+ to avoid quality degradation from re-encoding.
+
+ Args:
+ filepath: Path to the audio file
+ codec: Target codec - 'mp3', 'flac', or 'opus'
+ bitrate: Bitrate for lossy codecs (ignored for FLAC)
+
+ Returns:
+ tuple: (new_filepath, new_mime_type) or (original_filepath, None) if skipped
+ """
+ # Lossless formats that benefit from compression
+ lossless_extensions = {'.wav', '.aiff', '.aif'}
+
+ ext = os.path.splitext(filepath)[1].lower()
+
+ # Skip if not a lossless format
+ if ext not in lossless_extensions:
+ current_app.logger.debug(f"Skipping compression for {filepath} - not a lossless format")
+ return filepath, None
+
+ # Skip if target is same as source (e.g., FLAC to FLAC when source is already FLAC)
+ if ext == '.flac' and codec == 'flac':
+ current_app.logger.debug(f"Skipping compression for {filepath} - already FLAC")
+ return filepath, None
+
+ # Determine output extension and MIME type
+ codec_info = {
+ 'mp3': {'ext': '.mp3', 'mime': 'audio/mpeg'},
+ 'flac': {'ext': '.flac', 'mime': 'audio/flac'},
+ 'opus': {'ext': '.opus', 'mime': 'audio/opus'}
+ }
+
+ if codec not in codec_info:
+ current_app.logger.warning(f"Unknown codec '{codec}', defaulting to mp3")
+ codec = 'mp3'
+
+ output_ext = codec_info[codec]['ext']
+ output_mime = codec_info[codec]['mime']
+
+ base_filepath = os.path.splitext(filepath)[0]
+ temp_filepath = f"{base_filepath}_compressed_temp{output_ext}"
+ final_filepath = f"{base_filepath}{output_ext}"
+
+ try:
+ # Get original file size for logging
+ original_size = os.path.getsize(filepath)
+
+ # Build FFmpeg command based on codec
+ if codec == 'mp3':
+ cmd = [
+ 'ffmpeg', '-i', filepath, '-y',
+ '-acodec', 'libmp3lame',
+ '-b:a', bitrate,
+ '-ac', '1', # Mono for speech
+ temp_filepath
+ ]
+ elif codec == 'flac':
+ cmd = [
+ 'ffmpeg', '-i', filepath, '-y',
+ '-acodec', 'flac',
+ '-compression_level', '8', # Maximum compression
+ temp_filepath
+ ]
+ elif codec == 'opus':
+ cmd = [
+ 'ffmpeg', '-i', filepath, '-y',
+ '-acodec', 'libopus',
+ '-b:a', bitrate,
+ temp_filepath
+ ]
+
+ current_app.logger.info(f"Compressing {filepath} to {codec.upper()}...")
+ subprocess.run(cmd, check=True, capture_output=True, text=True)
+
+ # Get compressed file size
+ compressed_size = os.path.getsize(temp_filepath)
+ ratio = (1 - compressed_size / original_size) * 100 if original_size > 0 else 0
+
+ current_app.logger.info(
+ f"Compressed {filepath}: {original_size / 1024 / 1024:.1f}MB -> "
+ f"{compressed_size / 1024 / 1024:.1f}MB ({ratio:.1f}% reduction)"
+ )
+
+ # Remove original and rename temp to final
+ os.remove(filepath)
+ os.rename(temp_filepath, final_filepath)
+
+ return final_filepath, output_mime
+
+ except subprocess.CalledProcessError as e:
+ current_app.logger.error(f"FFmpeg compression failed for {filepath}: {e.stderr}")
+ # Clean up temp file if it exists
+ if os.path.exists(temp_filepath):
+ os.remove(temp_filepath)
+ raise Exception(f"Audio compression failed: {e.stderr}")
+ except FileNotFoundError:
+ current_app.logger.error("FFmpeg command not found for compression.")
+ raise Exception("Audio conversion tool (FFmpeg) not found on server.")
+ except Exception as e:
+ current_app.logger.error(f"Error compressing audio {filepath}: {str(e)}")
+ # Clean up temp file if it exists
+ if os.path.exists(temp_filepath):
+ os.remove(temp_filepath)
+ raise
+
def transcribe_audio_asr(app_context, recording_id, filepath, original_filename, start_time, mime_type=None, language=None, diarize=False, min_speakers=None, max_speakers=None, tag_id=None):
"""Transcribes audio using the ASR webservice."""