diff --git a/apps/import/routes.py b/apps/import/routes.py
index c5bb80fdb..1a720ea69 100644
--- a/apps/import/routes.py
+++ b/apps/import/routes.py
@@ -13,6 +13,7 @@ from flask import Blueprint, jsonify, render_template, request
 from werkzeug.utils import secure_filename
 
 from convey import emit, state
+from media import MEDIA_EXTENSIONS
 from think.detect_created import detect_created
 from think.importers.utils import (
     build_import_info,
@@ -120,7 +121,7 @@ SOURCE_METADATA = [
         "input_type": "file",
         "upload_prompt": "Upload an audio file (.m4a, .mp3, .wav)",
         "has_guide": False,
-        "accept": ".m4a,.mp3,.wav,.ogg,.webm",
+        "accept": ",".join(sorted(MEDIA_EXTENSIONS)),
     },
     {
         "name": "document",
diff --git a/media.py b/media.py
new file mode 100644
index 000000000..7de600c87
--- /dev/null
+++ b/media.py
@@ -0,0 +1,25 @@
+# SPDX-License-Identifier: AGPL-3.0-only
+# Copyright (c) 2026 sol pbc
+
+"""Media format registry - single source of truth for extensions, MIME types, and kind."""
+
+FORMATS = [
+    (".flac", "audio/flac", "audio"),
+    (".opus", "audio/opus", "audio"),
+    (".ogg", "audio/ogg", "audio"),
+    (".m4a", "audio/mp4", "audio"),
+    (".mp3", "audio/mpeg", "audio"),
+    (".wav", "audio/wav", "audio"),
+    (".webm", "video/webm", "video"),
+    (".mp4", "video/mp4", "video"),
+    (".mov", "video/quicktime", "video"),
+]
+
+AUDIO_EXTENSIONS: frozenset[str] = frozenset(
+    ext for ext, _, kind in FORMATS if kind == "audio"
+)
+VIDEO_EXTENSIONS: frozenset[str] = frozenset(
+    ext for ext, _, kind in FORMATS if kind == "video"
+)
+MEDIA_EXTENSIONS: frozenset[str] = frozenset(ext for ext, _, _ in FORMATS)
+MIME_TYPES: dict[str, str] = {ext: mime for ext, mime, _ in FORMATS}
diff --git a/observe/transcribe/main.py b/observe/transcribe/main.py
index dfa8e2a35..053148294 100644
--- a/observe/transcribe/main.py
+++ b/observe/transcribe/main.py
@@ -54,6 +54,7 @@ from pathlib import Path
 
 import numpy as np
 
+from media import AUDIO_EXTENSIONS as SUPPORTED_AUDIO_FORMATS
 from observe.transcribe import (
     BACKEND_REGISTRY,
     get_backend,
@@ -99,9 +100,6 @@ MIN_STATEMENT_DURATION = 0.3
 # Number of recent entity names to load for transcription context
 ENTITY_NAMES_LIMIT = 40
 
-# Supported audio file formats for transcription
-SUPPORTED_AUDIO_FORMATS = {".flac", ".m4a", ".mp3", ".ogg", ".opus", ".wav"}
-
 # Module-level voice encoder cache
 _voice_encoder = None
 
diff --git a/observe/utils.py b/observe/utils.py
index 1020b0b3c..396390987 100644
--- a/observe/utils.py
+++ b/observe/utils.py
@@ -16,6 +16,8 @@ from pathlib import Path
 import numpy as np
 import soundfile as sf
 
+from media import AUDIO_EXTENSIONS as _AUDIO_EXTENSIONS
+from media import VIDEO_EXTENSIONS as _VIDEO_EXTENSIONS
 from think.utils import day_path
 
 logger = logging.getLogger(__name__)
@@ -23,8 +25,8 @@ logger = logging.getLogger(__name__)
 # Standard sample rate for audio processing
 SAMPLE_RATE = 16000
 
-VIDEO_EXTENSIONS = (".webm", ".mp4", ".mov")
-AUDIO_EXTENSIONS = (".flac", ".ogg", ".m4a", ".opus", ".mp3")
+VIDEO_EXTENSIONS = tuple(_VIDEO_EXTENSIONS)
+AUDIO_EXTENSIONS = tuple(_AUDIO_EXTENSIONS)
 
 
 def audio_to_flac_bytes(audio: np.ndarray, sample_rate: int) -> bytes:
diff --git a/pyproject.toml b/pyproject.toml
index c5358e525..1b3448431 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -104,7 +104,7 @@ Documentation = "https://github.com/solpbc/solstone/blob/main/README.md"
 include = ["apps*", "think*", "convey*", "observe*", "muse*"]
 
 [tool.setuptools]
-py-modules = ["sol"]
+py-modules = ["media", "sol"]
 
 [tool.setuptools.package-data]
 apps = ["*/templates/*.html", "*/muse/*.md"]
diff --git a/tests/test_retention.py b/tests/test_retention.py
index 449bec6a2..14614bda3 100644
--- a/tests/test_retention.py
+++ b/tests/test_retention.py
@@ -26,7 +26,7 @@ from think.retention import (
 
 class TestIsRawMedia:
     def test_audio_extensions(self, tmp_path):
-        for ext in (".flac", ".opus", ".ogg", ".m4a"):
+        for ext in (".flac", ".opus", ".ogg", ".m4a", ".mp3", ".wav"):
             p = tmp_path / f"audio{ext}"
             p.touch()
             assert is_raw_media(p), f"{ext} should be raw media"
diff --git a/think/importers/shared.py b/think/importers/shared.py
index b0cb0e8a3..ad8604444 100644
--- a/think/importers/shared.py
+++ b/think/importers/shared.py
@@ -12,6 +12,7 @@ import shutil
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Callable
 
+from media import MIME_TYPES
 from think.importers.utils import save_import_file, write_import_metadata
 from think.utils import day_path, get_journal, now_ms
 
@@ -283,15 +284,7 @@ def write_markdown_segments(
 
 # MIME type mapping for import metadata
 _MIME_TYPES = {
-    ".m4a": "audio/mp4",
-    ".mp3": "audio/mpeg",
-    ".wav": "audio/wav",
-    ".flac": "audio/flac",
-    ".ogg": "audio/ogg",
-    ".opus": "audio/opus",
-    ".mp4": "video/mp4",
-    ".webm": "video/webm",
-    ".mov": "video/quicktime",
+    **MIME_TYPES,
     ".txt": "text/plain",
     ".md": "text/markdown",
     ".pdf": "application/pdf",
diff --git a/think/retention.py b/think/retention.py
index 8eaf2bf71..e43622d76 100644
--- a/think/retention.py
+++ b/think/retention.py
@@ -23,6 +23,9 @@ from datetime import datetime
 from pathlib import Path
 from typing import Any
 
+from media import AUDIO_EXTENSIONS as RAW_AUDIO_EXTENSIONS
+from media import MEDIA_EXTENSIONS as RAW_MEDIA_EXTENSIONS
+from media import VIDEO_EXTENSIONS as RAW_VIDEO_EXTENSIONS
 from think.utils import day_dirs, get_journal, iter_segments
 
 logger = logging.getLogger(__name__)
@@ -31,15 +34,11 @@ logger = logging.getLogger(__name__)
 # Raw media file identification
 # ---------------------------------------------------------------------------
 
-RAW_AUDIO_EXTENSIONS = frozenset({".flac", ".opus", ".ogg", ".m4a"})
-RAW_VIDEO_EXTENSIONS = frozenset({".webm", ".mov", ".mp4"})
-RAW_MEDIA_EXTENSIONS = RAW_AUDIO_EXTENSIONS | RAW_VIDEO_EXTENSIONS
-
 
 def is_raw_media(path: Path) -> bool:
     """Check if a file is raw media (layer 1 capture).
 
-    Raw media: *.flac, *.opus, *.ogg, *.m4a (audio),
+    Raw media: *.flac, *.opus, *.ogg, *.m4a, *.mp3, *.wav (audio),
     *.webm, *.mov, *.mp4 (video), monitor_*_diff.png (screen diffs).
     """
     if path.suffix.lower() in RAW_MEDIA_EXTENSIONS:
diff --git a/think/utils.py b/think/utils.py
index 0854e7cc2..eb777202c 100644
--- a/think/utils.py
+++ b/think/utils.py
@@ -26,6 +26,8 @@ import platformdirs
 from dotenv import load_dotenv
 from timefhuman import timefhuman
 
+from media import MIME_TYPES
+
 DATE_RE = re.compile(r"\d{8}")
 _journal_path_cache: str | None = None
 
@@ -800,24 +802,8 @@ def get_raw_file(day: str, name: str) -> tuple[str, str, Any]:
     if not rel:
         raise ValueError(f"No 'raw' field found in metadata for {name}")
 
-    # Determine MIME type from raw file extension
-    if rel.endswith(".flac"):
-        mime = "audio/flac"
-    elif rel.endswith(".ogg"):
-        mime = "audio/ogg"
-    elif rel.endswith(".m4a"):
-        mime = "audio/mp4"
-    elif rel.endswith(".png"):
-        mime = "image/png"
-    elif rel.endswith(".webm"):
-        mime = "video/webm"
-    elif rel.endswith(".mp4"):
-        mime = "video/mp4"
-    elif rel.endswith(".mov"):
-        mime = "video/quicktime"
-    else:
-        # Default fallback for unknown types
-        mime = "application/octet-stream"
+    suffix = Path(rel).suffix.lower()
+    mime = {**MIME_TYPES, ".png": "image/png"}.get(suffix, "application/octet-stream")
 
     return rel, mime, meta