diff --git a/.changeset/audio-source-clear-in-flight.md b/.changeset/audio-source-clear-in-flight.md
new file mode 100644
index 00000000..9daccbfe
--- /dev/null
+++ b/.changeset/audio-source-clear-in-flight.md
@@ -0,0 +1,5 @@
+---
+"livekit-rtc": patch
+---
+
+`AudioSource.clear_queue` stops a `capture_frame` in flight: a frame longer than the queue no longer keeps playing after a clear, and a capture right after a clear no longer fails.
diff --git a/livekit-rtc/rust-sdks b/livekit-rtc/rust-sdks
index 7f3ddc08..54508c56 160000
--- a/livekit-rtc/rust-sdks
+++ b/livekit-rtc/rust-sdks
@@ -1 +1 @@
-Subproject commit 7f3ddc08bda60e318d4acdd1625b0723ac6c009f
+Subproject commit 54508c56f5875c1d84106bd843c481647e637d2e
diff --git a/livekit-rtc/tests/test_audio_source.py b/livekit-rtc/tests/test_audio_source.py
new file mode 100644
index 00000000..09a50184
--- /dev/null
+++ b/livekit-rtc/tests/test_audio_source.py
@@ -0,0 +1,63 @@
+# Copyright 2026 LiveKit, Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""Regression tests for AudioSource.clear_queue."""
+
+from __future__ import annotations
+
+import asyncio
+import time
+
+from livekit import rtc
+
+SAMPLE_RATE = 24000
+
+
+async def test_clear_queue_cuts_a_capture_longer_than_the_queue() -> None:
+    """Audio captured before clear_queue stops right after it.
+
+    A frame longer than the queue is fed to it chunk by chunk; clear_queue must also drop the
+    chunks not queued yet, or they keep playing after an interruption.
+    """
+    source = rtc.AudioSource(SAMPLE_RATE, 1, queue_size_ms=50)
+    track = rtc.LocalAudioTrack.create_audio_track("clear", source)
+    stream = rtc.AudioStream(track, sample_rate=SAMPLE_RATE, num_channels=1)
+    heard: list[float] = []
+
+    async def listen() -> None:
+        async for event in stream:
+            if any(event.frame.data):
+                heard.append(time.monotonic())
+
+    listener = asyncio.create_task(listen())
+    samples = SAMPLE_RATE // 2  # 500 ms, ten times the queue
+    frame = rtc.AudioFrame(b"\xe8\x03" * samples, SAMPLE_RATE, 1, samples)
+    try:
+        capture = asyncio.create_task(source.capture_frame(frame))
+        await asyncio.sleep(0.025)
+        cleared = time.monotonic()
+        source.clear_queue()
+        # the next reply starts at once, while the cut capture winds down
+        silence = SAMPLE_RATE // 100
+        await source.capture_frame(rtc.AudioFrame(b"\x00\x00" * silence, SAMPLE_RATE, 1, silence))
+        await asyncio.wait_for(capture, 0.1)
+        await asyncio.sleep(0.6)
+    finally:
+        listener.cancel()
+        await asyncio.gather(listener, return_exceptions=True)
+        await stream.aclose()
+        await source.aclose()
+
+    after_clear = max((t - cleared for t in heard), default=0.0)
+    assert after_clear < 0.02, f"audio played {after_clear * 1000:.0f} ms after clear_queue"