Silence detection

TODO.md

@@ -63,7 +63,7 @@ Changes for 2024 based on real world usage in 2023 and feedback
 - [x] Single app client lock (ie only ONE client per username/password)'
 - [ ] Way to parse filename into rotator, start/end date
   - [ ] Submit form built into Tomato?
-- [ ] Silence detection REJECTs audio assets in backend (if there's more than 2 seconds?) (behind FEATURE flag)
+- [x] Silence detection REJECTs audio assets in backend (if there's more than 2 seconds?) (behind FEATURE flag)
 - [x] Export all audio assets as zip
   - [x] Import as well (have to be careful with different `protocol.json:protocol_version`)
 - [ ] Add configurable silence between ads. Crossfade, with fade points? Fancy!

server/tomato/ffmpeg.py

@@ -3,6 +3,7 @@
 import json
 import logging
 import math
+import re
 import shlex
 import subprocess
 
@@ -11,6 +12,7 @@
 
 logger = logging.getLogger(__name__)
 FFProbe = namedtuple("FFProbe", ("format", "duration", "title"))
+SILENCEDETECT_RE = re.compile(r"^\[silencedetect.+silence_duration: ([\d\.]+)", re.MULTILINE)
 
 
 def run_command(args):
@@ -123,3 +125,34 @@ def ffmpeg_convert(infile, outfile):
         trimmed_wav_file.unlink(missing_ok=True)
 
     return cmd.returncode == 0
+
+
+def silence_detect(infile):
+    if config.REJECT_SILENCE_LENGTH == 0:
+        return (False, 0)
+
+    cmd = run_command(
+        (
+            "ffmpeg",
+            "-y",
+            "-i",
+            infile,
+            "-hide_banner",
+            "-af",
+            f"silencedetect=n=-32dB:d={config.REJECT_SILENCE_LENGTH}",
+            "-f",
+            "null",
+            "-",
+        ),
+    )
+    if cmd.returncode != 0:
+        logger.error(f"ffmpeg returned {cmd.returncode} while detecting silence in {infile}: {cmd.stderr}")
+        return (False, 0)
+
+    silences = SILENCEDETECT_RE.findall(cmd.stderr)
+    if silences:
+        max_silence = max(math.ceil(float(s)) for s in SILENCEDETECT_RE.findall(cmd.stderr))
+        logger.warning(f"{infile} has a silence of {max_silence} seconds in it!")
+        return (True, max_silence)
+    else:
+        return (False, 0)

server/tomato/models/asset.py

@@ -17,6 +17,7 @@
 from constance import config
 from dirtyfields import DirtyFieldsMixin
 
+from ..ffmpeg import silence_detect
 from ..utils import listdir_recursive
 from .base import (
     FILE_MAX_LENGTH,
@@ -87,22 +88,33 @@ def serialize(self):
 
     def full_clean(self, *args, **kwargs):
         super().full_clean(*args, **kwargs)
-        if config.PREVENT_DUPLICATE_ASSETS and self.file and "file" in self.get_dirty_fields():
-            md5sum = self.generate_md5sum()
-            querysets = {
-                Asset: Asset.objects.filter(pre_process_md5sum=md5sum),
-                AssetAlternate: AssetAlternate.objects.filter(pre_process_md5sum=md5sum),
-            }
-            if self.id is not None:
-                querysets[self._meta.model] = querysets[self._meta.model].exclude(id=self.id)
-            if any(qs.exists() for qs in querysets.values()):
-                raise ValidationError({
-                    "__all__": mark_safe(
-                        "An audio asset already exists with this audio file. Rejecting duplicate. You can turn this"
-                        " feature off with setting <code>PREVENT_DUPLICATES</code>."
-                    ),
-                    "file": "A duplicate of this file already exists.",
-                })
+        if self.file and "file" in self.get_dirty_fields():
+            if config.PREVENT_DUPLICATE_ASSETS:
+                md5sum = self.generate_md5sum()
+                querysets = {
+                    Asset: Asset.objects.filter(pre_process_md5sum=md5sum),
+                    AssetAlternate: AssetAlternate.objects.filter(pre_process_md5sum=md5sum),
+                }
+                if self.id is not None:
+                    querysets[self._meta.model] = querysets[self._meta.model].exclude(id=self.id)
+                if any(qs.exists() for qs in querysets.values()):
+                    raise ValidationError({
+                        "__all__": mark_safe(
+                            "An audio asset already exists with this audio file. Rejecting duplicate. You can turn this"
+                            " feature off with setting <code>PREVENT_DUPLICATES</code>."
+                        ),
+                        "file": "A duplicate of this file already exists.",
+                    })
+            if config.REJECT_SILENCE_LENGTH > 0:
+                has_silence, silence_duration = silence_detect(self.file.real_path)
+                if has_silence:
+                    raise ValidationError({
+                        "__all__": mark_safe(
+                            f"This audio asset contains a silence of {silence_duration}s. Rejecting. You feature off"
+                            " with setting <code>REJECT_SILENCE_LENGTH</code>."
+                        ),
+                        "file": f"This asset contains a silence of {silence_duration} seconds.",
+                    })
 
     def save(self, dont_overwrite_original_filename=False, *args, **kwargs):
         if not dont_overwrite_original_filename and "file" in self.get_dirty_fields():

server/tomato/settings.py

@@ -313,6 +313,14 @@ def validate_reset_times(values):
             "validators": (validate_reset_times,),
         },
     ),
+    "silence": (
+        "django.forms.IntegerField",
+        {
+            "widget": "django.forms.TextInput",
+            "min_value": 0,
+            "max_value": 60,
+        },
+    ),
 }
 
 CONSTANCE_CONFIG = {
@@ -439,6 +447,11 @@ def validate_reset_times(values):
     "WARN_ON_EMPTY_ROTATORS": (True, "Warn when a rotator is disabled or has no eligible assets to choose from."),
     "RELOAD_PLAYLIST_AFTER_DATA_CHANGES": (False, "Reload all connected client playlists when a data change occurs."),
     "ONE_CLIENT_LOGIN_PER_ACCOUNT": (False, "Only allow one desktop client to be connected per account."),
+    "REJECT_SILENCE_LENGTH": (
+        0,
+        mark_safe("Reject silence assets of this many seconds. <strong>Set to 0 disable</strong>."),
+        "silence",
+    ),
 }
 
 CONSTANCE_CONFIG_FIELDSETS = OrderedDict((
@@ -458,6 +471,7 @@ def validate_reset_times(values):
             "EXTRACT_METADATA_FROM_FILE",
             "PREVENT_DUPLICATE_ASSETS",
             "TRIM_SILENCE",
+            "REJECT_SILENCE_LENGTH",
             "RELOAD_PLAYLIST_AFTER_DATA_CHANGES",
         ),
     ),
@@ -483,7 +497,7 @@ def validate_reset_times(values):
     "from constance import config",
     "from user_messages import api as user_messages_api",
     "from tomato import constants",
-    "from tomato.ffmpeg import ffmpeg_convert, ffprobe",
+    "from tomato.ffmpeg import ffmpeg_convert, ffprobe, silence_detect",
     "from tomato.models import export_data_as_zip, import_data_from_zip, serialize_for_api",
     "from tomato.tasks import bulk_process_assets, process_asset, cleanup",
 ]