Write standard ID3 frames for TTS tags (#182819)

homeassistant/components/tts/__init__.py

@@ -18,7 +18,7 @@ from typing import Any, Final, Protocol
 
 from aiohttp import web
 import mutagen
-from mutagen.id3 import ID3, TextFrame as ID3Text
+from mutagen.id3 import TALB, TIT2, TPE1, Encoding
 import probatio
 from propcache.api import cached_property
 
@@ -1195,23 +1195,9 @@ class SpeechManager:
             if tts_file is not None:
                 if not tts_file.tags:
                     tts_file.add_tags()
-                if isinstance(tts_file.tags, ID3):
-                    tts_file["artist"] = ID3Text(
-                        encoding=3,
-                        text=artist,  # type: ignore[no-untyped-call]
-                    )
-                    tts_file["album"] = ID3Text(
-                        encoding=3,
-                        text=album,  # type: ignore[no-untyped-call]
-                    )
-                    tts_file["title"] = ID3Text(
-                        encoding=3,
-                        text=message,  # type: ignore[no-untyped-call]
-                    )
-                else:
-                    tts_file["artist"] = artist
-                    tts_file["album"] = album
-                    tts_file["title"] = message
+                tts_file.tags.add(TPE1(encoding=Encoding.UTF8, text=artist))  # type: ignore[no-untyped-call]
+                tts_file.tags.add(TALB(encoding=Encoding.UTF8, text=album))  # type: ignore[no-untyped-call]
+                tts_file.tags.add(TIT2(encoding=Encoding.UTF8, text=message))  # type: ignore[no-untyped-call]
                 data_bytes.seek(0)
                 tts_file.save(data_bytes)
         except mutagen.MutagenError as err:

tests/components/tts/test_init.py

@@ -11,6 +11,8 @@ from unittest.mock import AsyncMock, MagicMock, Mock, patch
 import wave
 
 from freezegun.api import FrozenDateTimeFactory
+import mutagen
+from mutagen.id3 import TIT2, TPE1, Encoding
 import pytest
 
 from homeassistant.components import ffmpeg, tts
@@ -2327,3 +2329,111 @@ def test_write_tags_keeps_single_id3_tag() -> None:
 
     assert tagged.startswith(b"ID3")
     assert tagged.count(b"ID3") == 1
+
+
+def test_write_tags_sets_standard_id3_frames() -> None:
+    """Test tagging audio carrying only the encoder frame sets standard frames."""
+    data = load_fixture_bytes("tagged.mp3", DOMAIN)
+    assert list(mutagen.File(io.BytesIO(data)).tags) == ["TSSE"]
+
+    tagged = ORIG_WRITE_TAGS(
+        "42f18378fd4393d18c8dd11d03fa9563c1e54491_en-us_-_test.mp3",
+        data,
+        "Test",
+        "There is someone at the door.",
+        "en",
+        None,
+    )
+
+    tags = mutagen.File(io.BytesIO(tagged)).tags
+    assert tags["TPE1"].text == ["en"]
+    assert tags["TALB"].text == ["Test"]
+    assert tags["TIT2"].text == ["There is someone at the door."]
+    assert "TSSE" in tags
+
+
+def test_write_tags_adds_tag_to_untagged_audio() -> None:
+    """Test audio arriving without an ID3 tag gets one holding the frames."""
+    data = load_fixture_bytes("untagged.mp3", DOMAIN)
+    assert not data.startswith(b"ID3")
+
+    tagged = ORIG_WRITE_TAGS(
+        "42f18378fd4393d18c8dd11d03fa9563c1e54491_en-us_-_test.mp3",
+        data,
+        "Test",
+        "There is someone at the door.",
+        "en",
+        None,
+    )
+
+    assert tagged.startswith(b"ID3")
+    assert tagged.count(b"ID3") == 1
+    tags = mutagen.File(io.BytesIO(tagged)).tags
+    assert tags["TPE1"].text == ["en"]
+    assert tags["TALB"].text == ["Test"]
+    assert tags["TIT2"].text == ["There is someone at the door."]
+
+
+def test_write_tags_overwrites_id3v1_metadata() -> None:
+    """Test audio arriving with only an ID3v1 trailer gets the frames rewritten."""
+    data = load_fixture_bytes("id3v1.mp3", DOMAIN)
+    assert not data.startswith(b"ID3")
+    assert data[-128:-125] == b"TAG"
+
+    tagged = ORIG_WRITE_TAGS(
+        "42f18378fd4393d18c8dd11d03fa9563c1e54491_en-us_-_test.mp3",
+        data,
+        "Test",
+        "There is someone at the door.",
+        "en",
+        None,
+    )
+
+    assert tagged.startswith(b"ID3")
+    tags = mutagen.File(io.BytesIO(tagged)).tags
+    assert tags["TPE1"].text == ["en"]
+    assert tags["TALB"].text == ["Test"]
+    assert tags["TIT2"].text == ["There is someone at the door."]
+
+
+def test_write_tags_replaces_existing_frames() -> None:
+    """Test frames carried through conversion from the provider are replaced."""
+    data = load_fixture_bytes("untagged.mp3", DOMAIN)
+    source = io.BytesIO(data)
+    source.name = "source.mp3"
+    source_file = mutagen.File(source)
+    source_file.add_tags()
+    source_file.tags.add(TIT2(encoding=Encoding.UTF8, text="Provider title"))
+    source_file.tags.add(TPE1(encoding=Encoding.UTF8, text="Provider artist"))
+    source.seek(0)
+    source_file.save(source)
+
+    tagged = ORIG_WRITE_TAGS(
+        "42f18378fd4393d18c8dd11d03fa9563c1e54491_en-us_-_test.mp3",
+        source.getvalue(),
+        "Test",
+        "There is someone at the door.",
+        "en",
+        None,
+    )
+
+    tags = mutagen.File(io.BytesIO(tagged)).tags
+    assert tags["TIT2"].text == ["There is someone at the door."]
+    assert tags["TPE1"].text == ["en"]
+
+
+def test_write_tags_uses_voice_as_artist() -> None:
+    """Test the voice option replaces the language as the artist frame."""
+    data = load_fixture_bytes("tagged.mp3", DOMAIN)
+
+    tagged = ORIG_WRITE_TAGS(
+        "42f18378fd4393d18c8dd11d03fa9563c1e54491_en-us_-_test.mp3",
+        data,
+        "Test",
+        "There is someone at the door.",
+        "en",
+        {"voice": "JennyNeural"},
+    )
+
+    tags = mutagen.File(io.BytesIO(tagged)).tags
+    assert tags["TPE1"].text == ["JennyNeural"]