Add RFC 6381 codec info BitDepth and PixelFormat to video-audit, bids-inject, and split-video

.ai/task-bids-inject.md

@@ -79,6 +79,9 @@ Tracks implementation progress against [spec-bids-inject.md](spec-bids-inject.md
 - [x] Media suffix determination (`_video` / `_audio` / `_audiovideo`) from `videos.tsv`
 - [x] Delegate to `split-video` Python API (`do_main`)
 - [x] Build `sidecar_metadata` dict from `record.metadata.TaskName` and pass to `do_main`
+- [x] Populate `VideoCodecRFC6381` / `AudioCodecRFC6381` in `sidecar_metadata` via `get_audio_video_info_ffprobe` (TODO: future — read from `videos.tsv` columns instead)
+- [x] Populate `BitDepth` / `PixelFormat` in `sidecar_metadata` from `VideoInfo.bit_depth` / `VideoInfo.pix_fmt`
+- [x] ffprobe errors logged and injection continues (RFC6381 / BitDepth / PixelFormat silently omitted)
 
 ### Dry-run mode
 - [x] Skip `split-video` call and file writes when `--dry-run`
@@ -210,6 +213,18 @@ Test file location: `tests/qr/test_bids_inject.py` (mirrors `tests/audio/test_au
 - [x] `_is_scans_file` — `*_scans.tsv` existing file → `True`
 - [x] `_is_scans_file` — directory or non-matching name → `False`
 
+### sidecar_metadata propagation
+
+- [x] `_call_split_video` passes `sidecar_metadata` with `TaskName` to `split-video`
+- [x] `_call_split_video` passes empty `sidecar_metadata` when `TaskName` absent
+- [x] `_call_split_video` adds `VideoCodecRFC6381` / `AudioCodecRFC6381` to `sidecar_metadata` via ffprobe
+- [x] `_call_split_video` adds `BitDepth` / `PixelFormat` to `sidecar_metadata` via ffprobe
+- [x] ffprobe failure logs error and injection continues; RFC6381/BitDepth/PixelFormat silently omitted
+- [x] `_to_bids_model` uses `VideoCodecRFC6381` / `AudioCodecRFC6381` from `sidecar_metadata` when provided
+- [x] `_to_bids_model` defaults `VideoCodecRFC6381` / `AudioCodecRFC6381` to `"n/a"` when absent
+- [x] `_to_bids_model` writes `BitDepth` (int) / `PixelFormat` (str) from `sidecar_metadata` when present
+- [x] `_to_bids_model` omits `BitDepth` / `PixelFormat` when absent from `sidecar_metadata`
+
 ### Integration tests (with synthetic BIDS fixture)
 
 > Requires a small synthetic BIDS dataset fixture (a few `_scans.tsv` files, stub JSON

.ai/task-video-audit.md

@@ -82,6 +82,11 @@ Tracks implementation progress against [spec-video-audit.md](spec-video-audit.md
 - [x] `VaContext` — Pydantic model carrying all processing options
 - [x] `VaMode` — enum of operation modes
 - [x] `VaSource` — enum of audit sources
+- [x] `AudioInfo` — Pydantic model for audio stream info (codec, codec_long, codec_rfc6381, profile, sample_rate, channels, bits_per_sample, duration_sec, start_time, tag_str)
+- [x] `VideoInfo` — Pydantic model for video stream info (codec, codec_long, codec_rfc6381, profile, level, width, height, pix_fmt, bit_depth, fps, duration_sec, start_time, tag_str)
+- [x] `audio_codec_to_rfc6381` — RFC 6381 string for audio codec
+- [x] `video_codec_to_rfc6381` — RFC 6381 string for video codec
+- [x] `get_audio_video_info_ffprobe` — single ffprobe call returning `Tuple[AudioInfo, VideoInfo]`
 - [x] `do_audit_file` — audit a single video file (INTERNAL)
 - [x] `do_audit_dir` — audit all videos in a directory
 - [x] `do_audit_internal` — entry point for INTERNAL source
@@ -128,12 +133,27 @@ Test file location: `tests/qr/test_video_audit.py`
 - [x] `ffprobe` available (subprocess mock) → `True`
 - [x] `ffprobe` not found (`FileNotFoundError` mock) → `False`
 
-#### `get_audio_info_ffprobe`
-- [x] Success with DURATION tag → codec/sample_rate/channels/bits_per_sample/duration set
+#### `audio_codec_to_rfc6381`
+- [x] AAC LC → `"mp4a.40.2"`
+- [x] AAC HE-AAC → `"mp4a.40.5"`
+- [x] AAC unknown/None profile → defaults to `"mp4a.40.2"`
+- [x] MP3 → `"mp4a.69"`
+- [x] Opus → `"opus"`
+- [x] Unknown codec → `None`
+
+#### `video_codec_to_rfc6381`
+- [x] H.264 High@L4.2 → `"avc1.64002A"`
+- [x] H.264 Baseline@L3.0 → `"avc1.42001E"`
+- [x] H.264 None level → `"avc1.4D0000"`
+- [x] Unknown codec → `None`
+
+#### `get_audio_video_info_ffprobe` (replaces `get_audio_info_ffprobe`)
+- [x] Success — audio fields: codec/codec_long/profile/sample_rate/channels/start_time/tag_str/codec_rfc6381/duration set
+- [x] Success — video fields: codec/codec_long/profile/level/width/height/pix_fmt/bit_depth/fps/codec_rfc6381/duration/start_time/tag_str set
 - [x] Duration read from stream `duration` field (no DURATION tag)
-- [x] No audio streams → all fields `None`
-- [x] `FileNotFoundError` → returns empty `AudioInfo`
-- [x] `CalledProcessError` → returns empty `AudioInfo`
+- [x] No audio streams → AudioInfo all `None`, VideoInfo populated
+- [x] `FileNotFoundError` → returns empty AudioInfo and VideoInfo
+- [x] `CalledProcessError` → returns empty AudioInfo and VideoInfo
 
 #### `_compare_rec_ts`
 - [x] Both `n/a` → `0`

src/reprostim/qr/bids_inject.py

@@ -23,7 +23,10 @@
 
 from pydantic import BaseModel, Field
 
-from reprostim.qr.video_audit import find_video_audit_by_timerange
+from reprostim.qr.video_audit import (
+    find_video_audit_by_timerange,
+    get_audio_video_info_ffprobe,
+)
 
 # initialize the logger
 logger = logging.getLogger(__name__)
@@ -892,6 +895,21 @@ def _capturing_out_func(msg: str) -> None:
     if record.metadata and record.metadata.TaskName:
         sidecar_metadata["TaskName"] = record.metadata.TaskName
 
+    # TODO: in the future, read these fields from dedicated columns in videos.tsv
+    #       populated by video-audit, avoiding the extra ffprobe call here.
+    try:
+        ai, vi = get_audio_video_info_ffprobe(input_path)
+        if vi.codec_rfc6381:
+            sidecar_metadata["VideoCodecRFC6381"] = vi.codec_rfc6381
+        if ai.codec_rfc6381:
+            sidecar_metadata["AudioCodecRFC6381"] = ai.codec_rfc6381
+        if vi.bit_depth is not None:
+            sidecar_metadata["BitDepth"] = vi.bit_depth
+        if vi.pix_fmt:
+            sidecar_metadata["PixelFormat"] = vi.pix_fmt
+    except Exception as e:
+        logger.error(f"Failed to get video info for {input_path}: {e}")
+
     ret, split_results = split_video_main(
         input_path=input_path,
         output_path=output_path,

src/reprostim/qr/split_video.py

@@ -113,7 +113,9 @@ def _to_bids_model(sr: "SplitResult", sidecar_metadata: dict | None = None) -> d
 
     :param sr: SplitResult to convert
     :param sidecar_metadata: Optional dict with extra BIDS fields to inject.
-        Currently supports ``TaskName`` (written as the first field when present).
+        Supports ``TaskName`` (written as the first field when present),
+        ``VideoCodecRFC6381``, ``AudioCodecRFC6381``, ``BitDepth``, and
+        ``PixelFormat``.
     :return: Dict with BIDS-compliant metadata field names
     """
     result = {}
@@ -134,7 +136,9 @@ def _to_bids_model(sr: "SplitResult", sidecar_metadata: dict | None = None) -> d
 
     if sr.video_codec != "n/a":
         result["VideoCodec"] = sr.video_codec
-        result["VideoCodecRFC6381"] = "n/a"
+        result["VideoCodecRFC6381"] = (sidecar_metadata or {}).get(
+            "VideoCodecRFC6381", "n/a"
+        )
 
     if sr.video_frame_rate is not None:
         result["FrameRate"] = sr.video_frame_rate
@@ -151,9 +155,19 @@ def _to_bids_model(sr: "SplitResult", sidecar_metadata: dict | None = None) -> d
         except (ValueError, TypeError):
             pass
 
+    if sidecar_metadata:
+        bit_depth = sidecar_metadata.get("BitDepth")
+        if bit_depth is not None:
+            result["BitDepth"] = int(bit_depth)
+        pix_fmt = sidecar_metadata.get("PixelFormat")
+        if pix_fmt:
+            result["PixelFormat"] = pix_fmt
+
     if sr.audio_codec != "n/a":
         result["AudioCodec"] = sr.audio_codec
-        result["AudioCodecRFC6381"] = "n/a"
+        result["AudioCodecRFC6381"] = (sidecar_metadata or {}).get(
+            "AudioCodecRFC6381", "n/a"
+        )
 
     if sr.audio_sample_rate != "n/a":
         try: