diff --git a/server/reflector/hatchet/workflows/diarization_pipeline.py b/server/reflector/hatchet/workflows/diarization_pipeline.py
index 02668497..08f88554 100644
--- a/server/reflector/hatchet/workflows/diarization_pipeline.py
+++ b/server/reflector/hatchet/workflows/diarization_pipeline.py
@@ -48,6 +48,7 @@ from reflector.pipelines import topic_processing
 from reflector.processors import AudioFileWriterProcessor
 from reflector.processors.types import (
     TitleSummary,
+    TitleSummaryWithId,
     Word,
 )
 from reflector.processors.types import (
@@ -320,7 +321,6 @@ async def process_tracks(input: PipelineInput, ctx: Context) -> ProcessTracksRes
 
     target_language = participants_result.target_language
 
-    # Collect results from each track (don't mutate lists while iterating)
     track_words = []
     padded_tracks = []
     created_padded_files = set()
@@ -372,6 +372,7 @@ async def mixdown_tracks(input: PipelineInput, ctx: Context) -> MixdownResult:
     track_result = ctx.task_output(process_tracks)
     padded_tracks = track_result.padded_tracks
 
+    # TODO think of NonEmpty type to avoid those checks, e.g. sized.NonEmpty from https://github.com/antonagestam/phantom-types/
     if not padded_tracks:
         raise ValueError("No padded tracks to mixdown")
 
@@ -399,10 +400,10 @@ async def mixdown_tracks(input: PipelineInput, ctx: Context) -> MixdownResult:
         raise ValueError("No decodable audio frames in any track")
 
     output_path = tempfile.mktemp(suffix=".mp3")
-    duration_ms = [0.0]  # Mutable container for callback capture
+    duration_ms_callback_capture_container = [0.0]
 
     async def capture_duration(d):
-        duration_ms[0] = d
+        duration_ms_callback_capture_container[0] = d
 
     writer = AudioFileWriterProcessor(path=output_path, on_duration=capture_duration)
 
@@ -436,7 +437,7 @@ async def mixdown_tracks(input: PipelineInput, ctx: Context) -> MixdownResult:
 
     return MixdownResult(
         audio_key=storage_path,
-        duration=duration_ms[0],
+        duration=duration_ms_callback_capture_container[0],
         tracks_mixed=len(valid_urls),
     )
 
@@ -530,9 +531,6 @@ async def detect_topics(input: PipelineInput, ctx: Context) -> TopicsResult:
         TranscriptTopic,
         transcripts_controller,
     )
-    from reflector.processors.types import (  # noqa: PLC0415
-        TitleSummaryWithId as TitleSummaryWithIdProcessorType,
-    )
 
     word_objects = [Word(**w) for w in words]
     transcript_type = TranscriptType(words=word_objects)
@@ -550,7 +548,9 @@ async def detect_topics(input: PipelineInput, ctx: Context) -> TopicsResult:
                 transcript=data.transcript.text,
                 words=data.transcript.words,
             )
-            if isinstance(data, TitleSummaryWithIdProcessorType):
+            if isinstance(
+                data, TitleSummaryWithId
+            ):  # Celery parity: main_live_pipeline.py
                 topic.id = data.id
             await transcripts_controller.upsert_topic(transcript, topic)
             await append_event_and_broadcast(
@@ -720,10 +720,6 @@ async def finalize(input: PipelineInput, ctx: Context) -> FinalizeResult:
             TranscriptText,
             transcripts_controller,
         )
-        from reflector.processors.types import (  # noqa: PLC0415
-            Transcript as TranscriptType,
-        )
-        from reflector.processors.types import Word  # noqa: PLC0415
 
         transcript = await transcripts_controller.get_by_id(input.transcript_id)
         if transcript is None: