Include WAV recordings in mixed-format Babel manifests

lhotse/recipes/babel.py

@@ -5,6 +5,7 @@
 It has about 25 languages with 40h - 160h of training recordings and ~20h
 of development set recordings.
 """
+
 import logging
 import re
 from collections import defaultdict
@@ -21,7 +22,6 @@
     SupervisionSet,
     validate_recordings_and_supervisions,
 )
-from lhotse.manipulation import combine
 from lhotse.qa import (
     remove_missing_recordings_and_supervisions,
     trim_supervisions_to_recordings,
@@ -77,7 +77,8 @@ def prepare_single_babel_language(
             if there is more than once, it picks the first one (and emits a warning)
         - then, it will try to find `dev`, `eval`, and `training` splits inside
             (if any of them is not present, it will skip it with a warning)
-        - finally, it scans the selected location for SPHERE audio files and transcripts.
+        - finally, it scans the selected location for SPHERE and WAV audio files
+            and transcripts.
 
     :param corpus_dir: Path to the root of the LDC package with a BABEL language.
     :param output_dir: Path where the manifests are stored.json
@@ -107,13 +108,11 @@ def prepare_single_babel_language(
 
     for split in ("dev", "eval", "training"):
         audio_dir = corpus_dir / f"conversational/{split}/audio"
-        sph_recordings = RecordingSet.from_recordings(
-            Recording.from_file(p) for p in audio_dir.glob("*.sph")
-        )
-        wav_recordings = RecordingSet.from_recordings(
-            Recording.from_file(p) for p in audio_dir.glob("*.wav")
+        recordings = RecordingSet.from_recordings(
+            Recording.from_file(p)
+            for pattern in ("*.sph", "*.wav")
+            for p in audio_dir.glob(pattern)
         )
-        recordings = combine(sph_recordings, wav_recordings)
         if len(recordings) == 0:
             if split == "eval" and no_eval_ok:
                 continue
@@ -137,8 +136,8 @@ def prepare_single_babel_language(
             lines = p.read_text().splitlines() + [""]
             lines = [
                 prev_l
-                for prev_l, l in sliding_window(2, lines)
-                if not (prev_l.startswith("[") and l.startswith("["))
+                for prev_l, current_l in sliding_window(2, lines)
+                if not (prev_l.startswith("[") and current_l.startswith("["))
             ]
             # Add a None at the end so that the last timestamp is only used as "next_timestamp"
             # and ends the iretation (otherwise we'd lose the last segment).

test/recipes/test_babel.py

@@ -0,0 +1,44 @@
+from pathlib import Path
+
+from lhotse import AudioSource, Recording
+from lhotse.recipes.babel import prepare_single_babel_language
+
+
+def test_prepare_single_babel_language_includes_sph_and_wav(monkeypatch, tmp_path):
+    stems_and_suffixes = (
+        ("BABEL_BP_401_10001_20111024_205740_inLine", ".sph"),
+        ("BABEL_BP_401_10002_20111024_205740_inLine", ".wav"),
+    )
+    for split in ("dev", "training"):
+        audio_dir = tmp_path / "conversational" / split / "audio"
+        text_dir = tmp_path / "conversational" / split / "transcription"
+        audio_dir.mkdir(parents=True)
+        text_dir.mkdir(parents=True)
+
+        for stem, suffix in stems_and_suffixes:
+            audio_path = audio_dir / f"{stem}{suffix}"
+            audio_path.touch()
+            (text_dir / f"{stem}.txt").write_text("[0.0]\nhello\n[1.0]\n")
+
+    def fake_recording_from_file(path):
+        path = Path(path)
+        return Recording(
+            id=path.stem,
+            sources=[AudioSource(type="file", channels=[0], source=str(path))],
+            sampling_rate=8000,
+            num_samples=8000,
+            duration=1.0,
+        )
+
+    monkeypatch.setattr(Recording, "from_file", staticmethod(fake_recording_from_file))
+
+    manifests = prepare_single_babel_language(tmp_path, no_eval_ok=True)
+
+    for split in ("dev", "training"):
+        recordings = manifests[split]["recordings"]
+        assert not recordings.is_lazy
+        assert set(recordings.ids) == {stem for stem, _ in stems_and_suffixes}
+        assert {Path(r.sources[0].source).suffix for r in recordings} == {
+            ".sph",
+            ".wav",
+        }