🧱 Include video in reencoded preview media

frontend/src/editor/player.tsx

@@ -35,13 +35,16 @@ export function PlayerBar({ documentId, editor }: { documentId: string; editor:
   const [playbackRate, setPlaybackRate] = useLocalStorage('playbackRate', 1);
 
   const sources = useMemo(() => {
-    const mappedFiles =
-      data?.media_files.map((media) => {
-        return {
-          src: media.url,
-          type: media.content_type,
-        };
-      }) || [];
+    // do not play the original file, it may be large
+    const relevantMediaFiles =
+      data?.media_files.filter((media) => !media.tags.includes('original')) || [];
+
+    const mappedFiles = relevantMediaFiles.map((media) => {
+      return {
+        src: media.url,
+        type: media.content_type,
+      };
+    });
 
     return sortMediaFiles(mappedFiles);
   }, [data?.media_files]);

worker/transcribee_worker/config.py

@@ -17,10 +17,13 @@ class Settings(BaseSettings):
             "audio_bitrate": "128k",
             "ac": "1",
         },
-        "m4a": {
+        "video:mp4": {
             "format": "mp4",
             "audio_bitrate": "128k",
             "ac": "1",
+            "c:v": "libx264",
+            "crf": "26",
+            "preset": "faster",
         },
     }
 

worker/transcribee_worker/reencode.py

@@ -10,27 +10,37 @@ def get_duration(input_path: Path):
     return float(ffmpeg.probe(input_path)["format"]["duration"])
 
 
+def has_video(input_path: Path):
+    streams = ffmpeg.probe(input_path)["streams"]
+    for stream in streams:
+        if stream["codec_type"] == "video":
+            return True
+    return False
+
+
 async def reencode(
     input_path: Path,
     output_path: Path,
     output_params: dict[str, str],
     progress_callback: ProgressCallbackType,
     duration: float,
+    include_video: bool,
 ):
     def work(_):
-        cmd: subprocess.Popen = (
-            ffmpeg.input(input_path)
-            .output(
-                filename=output_path,
-                map="0:a",
-                loglevel="quiet",
-                stats=None,
-                progress="-",
-                map_metadata="-1",
-                **output_params
-            )
-            .run_async(pipe_stdout=True)
-        )
+        pipeline = ffmpeg.input(input_path)
+        streams = [pipeline.audio]
+        if include_video and has_video(input_path):
+            streams.append(pipeline.video)
+
+        cmd: subprocess.Popen = ffmpeg.output(
+            *streams,
+            filename=output_path,
+            loglevel="quiet",
+            stats=None,
+            progress="-",
+            map_metadata="-1",
+            **output_params
+        ).run_async(pipe_stdout=True)
         assert cmd.stdout
         raw_line: bytes
         progress_dict = {}

worker/transcribee_worker/worker.py

@@ -257,7 +257,7 @@ async def reencode(
 
         n_profiles = len(settings.REENCODE_PROFILES)
         for i, (profile, parameters) in enumerate(settings.REENCODE_PROFILES.items()):
-            output_path = self._get_tmpfile(f"reencode_{profile}")
+            output_path = self._get_tmpfile(f"reencode_{profile.replace(':', '_')}")
 
             await reencode(
                 document_audio,
@@ -269,6 +269,7 @@ async def reencode(
                     **kwargs,
                 ),
                 duration,
+                include_video=(profile.startswith("video:")),
             )
 
             tags = [f"profile:{profile}"] + [f"{k}:{v}" for k, v in parameters.items()]