diff --git a/lhotse/audio/backend.py b/lhotse/audio/backend.py
index 64a9e1449..562387b56 100644
--- a/lhotse/audio/backend.py
+++ b/lhotse/audio/backend.py
@@ -478,11 +478,7 @@ def torchaudio_supports_ffmpeg() -> bool:
     # If user has disabled ffmpeg-torchaudio, we don't need to check the version.
     if not _FFMPEG_TORCHAUDIO_INFO_ENABLED:
         return False
-
-    import torchaudio
-    from packaging import version
-
-    return version.parse(torchaudio.__version__) >= version.parse("0.12.0")
+    return check_torchaudio_version_gt("0.12.0")
 
 
 @lru_cache(maxsize=1)
@@ -494,14 +490,10 @@ def torchaudio_2_0_ffmpeg_enabled() -> bool:
     if not is_torchaudio_available():
         return False
 
-    import torchaudio
-    from packaging import version
-
-    ver = version.parse(torchaudio.__version__)
-    if ver >= version.parse("2.1.0"):
+    if check_torchaudio_version_gt("2.1.0"):
         # Enabled by default, disable with TORCHAUDIO_USE_BACKEND_DISPATCHER=0
         return os.environ.get("TORCHAUDIO_USE_BACKEND_DISPATCHER", "1") == "1"
-    if ver >= version.parse("2.0"):
+    if check_torchaudio_version_gt("2.0"):
         # Disabled by default, enable with TORCHAUDIO_USE_BACKEND_DISPATCHER=1
         return os.environ.get("TORCHAUDIO_USE_BACKEND_DISPATCHER", "0") == "1"
     return False
@@ -513,10 +505,14 @@ def torchaudio_soundfile_supports_format() -> bool:
     Returns ``True`` when torchaudio version is at least 0.9.0, which
     has support for ``format`` keyword arg in ``torchaudio.save()``.
     """
+    return check_torchaudio_version_gt("0.9.0")
+
+
+def check_torchaudio_version_gt(version: str) -> bool:
     import torchaudio
-    from packaging import version
+    from packaging import version as _version
 
-    return version.parse(torchaudio.__version__) >= version.parse("0.9.0")
+    return _version.parse(torchaudio.__version__) >= _version.parse(version)
 
 
 def torchaudio_info(
diff --git a/lhotse/shar/writers/audio.py b/lhotse/shar/writers/audio.py
index 69db168a6..91b4d1291 100644
--- a/lhotse/shar/writers/audio.py
+++ b/lhotse/shar/writers/audio.py
@@ -9,6 +9,8 @@
 from typing_extensions import Literal
 
 from lhotse import Recording
+from lhotse.audio.backend import check_torchaudio_version_gt
+from lhotse.augmentation import get_or_create_resampler
 from lhotse.shar.utils import to_shar_placeholder
 from lhotse.shar.writers.tar import TarWriter
 
@@ -53,6 +55,10 @@ def __init__(
             self.save_fn = partial(
                 torchaudio.backend.soundfile_backend.save, bits_per_sample=16
             )
+        if self.format == "opus":
+            assert check_torchaudio_version_gt(
+                "2.1.0"
+            ), "Writing OPUS files into Lhotse Shar requires torchaudio >= 2.1.0"
 
     def __enter__(self):
         self.tar_writer.__enter__()
@@ -79,19 +85,28 @@ def write(
         sampling_rate: int,
         manifest: Recording,
     ) -> None:
+        # Resampling is required for some versions of OPUS encoders.
+        # First resample the manifest which only adjusts the metadata;
+        # then resample the audio array to 48kHz.
+        value = torch.from_numpy(value)
+        if self.format == "opus" and sampling_rate != OPUS_DEFAULT_SAMPLING_RATE:
+            manifest = manifest.resample(OPUS_DEFAULT_SAMPLING_RATE)
+            value = get_or_create_resampler(sampling_rate, OPUS_DEFAULT_SAMPLING_RATE)(
+                value
+            )
+            sampling_rate = OPUS_DEFAULT_SAMPLING_RATE
+
         # Write binary data
         stream = BytesIO()
         self.save_fn(
             stream,
-            torch.from_numpy(value),
+            value,
             sampling_rate,
             format=self.format,
         )
         self.tar_writer.write(f"{key}.{self.format}", stream)
 
         # Write text manifest afterwards
-        if self.format == "opus" and sampling_rate != OPUS_DEFAULT_SAMPLING_RATE:
-            manifest = manifest.resample(OPUS_DEFAULT_SAMPLING_RATE)
         manifest = to_shar_placeholder(manifest)
         json_stream = BytesIO()
         print(
diff --git a/test/shar/test_write.py b/test/shar/test_write.py
index 1672b68b3..35a3f4487 100644
--- a/test/shar/test_write.py
+++ b/test/shar/test_write.py
@@ -6,6 +6,7 @@
 import pytest
 
 from lhotse import CutSet
+from lhotse.audio.backend import check_torchaudio_version_gt
 from lhotse.lazy import LazyJsonlIterator
 from lhotse.shar import AudioTarWriter, SharWriter, TarIterator, TarWriter
 from lhotse.testing.dummies import DummyManifest
@@ -65,7 +66,27 @@ def test_tar_writer_pipe(tmp_path: Path):
         assert f2.read() == b"test"
 
 
-@pytest.mark.parametrize("format", ["wav", "flac", "mp3", "opus"])
+@pytest.mark.parametrize(
+    "format",
+    [
+        "wav",
+        pytest.param(
+            "flac",
+            marks=pytest.mark.skipif(
+                not check_torchaudio_version_gt("0.12.1"),
+                reason="Torchaudio v0.12.1 or greater is required.",
+            ),
+        ),
+        # "mp3",  # apparently doesn't work in CI, mp3 encoder is missing
+        pytest.param(
+            "opus",
+            marks=pytest.mark.skipif(
+                not check_torchaudio_version_gt("2.1.0"),
+                reason="Torchaudio v2.1.0 or greater is required.",
+            ),
+        ),
+    ],
+)
 def test_audio_tar_writer(tmp_path: Path, format: str):
     from lhotse.testing.dummies import dummy_recording