diff --git a/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/shared/audio_wrappers.py b/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/shared/audio_wrappers.py
new file mode 100644
index 0000000000..978a1b42db
--- /dev/null
+++ b/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/shared/audio_wrappers.py
@@ -0,0 +1,247 @@
+import logging
+import time
+
+from opentelemetry import context as context_api
+from opentelemetry.instrumentation.openai.shared import (
+    _set_client_attributes,
+    _set_request_attributes,
+    _set_response_attributes,
+    _set_span_attribute,
+    metric_shared_attributes,
+    model_as_dict,
+)
+from opentelemetry.instrumentation.openai.utils import (
+    _with_audio_telemetry_wrapper,
+    dont_throw,
+    is_openai_v1,
+    start_as_current_span_async,
+)
+from opentelemetry.instrumentation.utils import _SUPPRESS_INSTRUMENTATION_KEY
+from opentelemetry.metrics import Counter, Histogram
+from opentelemetry.semconv.attributes.error_attributes import ERROR_TYPE
+from opentelemetry.semconv_ai import (
+    SUPPRESS_LANGUAGE_MODEL_INSTRUMENTATION_KEY,
+    SpanAttributes,
+)
+from opentelemetry.trace import SpanKind, Status, StatusCode
+
+SPAN_NAME = "openai.audio.transcriptions"
+
+logger = logging.getLogger(__name__)
+
+
+def _get_audio_duration(file):
+    """
+    Extract audio duration from file object.
+    Returns duration in seconds, or None if unable to determine.
+    """
+    try:
+        # Try to get duration from common audio libraries
+        # First check if it's a file-like object with a name attribute
+        if hasattr(file, "name"):
+            file_path = file.name
+        elif isinstance(file, (str, bytes)):
+            # If it's a path string or bytes
+            return None
+        else:
+            # If it's a file-like object without name, we can't easily determine duration
+            return None
+
+        # Try mutagen (supports many formats)
+        try:
+            from mutagen import File as MutagenFile
+
+            audio = MutagenFile(file_path)
+            if audio and hasattr(audio.info, "length"):
+                return audio.info.length
+        except (ImportError, Exception):
+            pass
+
+    except Exception as e:
+        logger.debug(f"Unable to extract audio duration: {e}")
+
+    return None
+
+
+@_with_audio_telemetry_wrapper
+def transcription_wrapper(
+    tracer,
+    duration_histogram: Histogram,
+    exception_counter: Counter,
+    wrapped,
+    instance,
+    args,
+    kwargs,
+):
+    if context_api.get_value(_SUPPRESS_INSTRUMENTATION_KEY) or context_api.get_value(
+        SUPPRESS_LANGUAGE_MODEL_INSTRUMENTATION_KEY
+    ):
+        return wrapped(*args, **kwargs)
+
+    with tracer.start_as_current_span(
+        name=SPAN_NAME,
+        kind=SpanKind.CLIENT,
+    ) as span:
+        _handle_request(span, kwargs, instance)
+
+        try:
+            # record time for duration
+            start_time = time.time()
+            response = wrapped(*args, **kwargs)
+            end_time = time.time()
+        except Exception as e:  # pylint: disable=broad-except
+            end_time = time.time()
+            duration = end_time - start_time if "start_time" in locals() else 0
+            attributes = {
+                "error.type": e.__class__.__name__,
+            }
+
+            # if there are legal duration, record it
+            if duration > 0 and duration_histogram:
+                duration_histogram.record(duration, attributes=attributes)
+            if exception_counter:
+                exception_counter.add(1, attributes=attributes)
+
+            span.set_attribute(ERROR_TYPE, e.__class__.__name__)
+            span.record_exception(e)
+            span.set_status(Status(StatusCode.ERROR, str(e)))
+            span.end()
+
+            raise
+
+        duration = end_time - start_time
+
+        _handle_response(
+            response,
+            span,
+            instance,
+            duration_histogram,
+            duration,
+        )
+
+        return response
+
+
+@_with_audio_telemetry_wrapper
+async def atranscription_wrapper(
+    tracer,
+    duration_histogram: Histogram,
+    exception_counter: Counter,
+    wrapped,
+    instance,
+    args,
+    kwargs,
+):
+    if context_api.get_value(_SUPPRESS_INSTRUMENTATION_KEY) or context_api.get_value(
+        SUPPRESS_LANGUAGE_MODEL_INSTRUMENTATION_KEY
+    ):
+        return await wrapped(*args, **kwargs)
+
+    async with start_as_current_span_async(
+        tracer=tracer,
+        name=SPAN_NAME,
+        kind=SpanKind.CLIENT,
+    ) as span:
+        _handle_request(span, kwargs, instance)
+
+        try:
+            # record time for duration
+            start_time = time.time()
+            response = await wrapped(*args, **kwargs)
+            end_time = time.time()
+        except Exception as e:  # pylint: disable=broad-except
+            end_time = time.time()
+            duration = end_time - start_time if "start_time" in locals() else 0
+            attributes = {
+                "error.type": e.__class__.__name__,
+            }
+
+            # if there are legal duration, record it
+            if duration > 0 and duration_histogram:
+                duration_histogram.record(duration, attributes=attributes)
+            if exception_counter:
+                exception_counter.add(1, attributes=attributes)
+
+            span.set_attribute(ERROR_TYPE, e.__class__.__name__)
+            span.record_exception(e)
+            span.set_status(Status(StatusCode.ERROR, str(e)))
+            span.end()
+
+            raise
+
+        duration = end_time - start_time
+
+        _handle_response(
+            response,
+            span,
+            instance,
+            duration_histogram,
+            duration,
+        )
+
+        return response
+
+
+@dont_throw
+def _handle_request(span, kwargs, instance):
+    _set_request_attributes(span, kwargs, instance)
+    _set_client_attributes(span, instance)
+
+    # Extract and set audio duration
+    file_param = kwargs.get("file")
+    if file_param:
+        audio_duration = _get_audio_duration(file_param)
+        if audio_duration is not None:
+            # _set_span_attribute(
+            #     span, SpanAttributes.LLM_OPENAI_AUDIO_INPUT_DURATION_SECONDS, audio_duration
+            # )
+            # TODO(Ata): come back here later when semconv is published
+            _set_span_attribute(
+                span, 'gen_ai.openai.audio.input.duration_seconds', audio_duration
+            )
+        else:
+            print("REMOVE ME : ATA-DBG : COULD NOT READ AUDIO FILE WITH MUTAGEN")
+
+
+@dont_throw
+def _handle_response(
+    response,
+    span,
+    instance=None,
+    duration_histogram=None,
+    duration=None,
+):
+    if is_openai_v1():
+        response_dict = model_as_dict(response)
+    else:
+        response_dict = response
+
+    # metrics record
+    _set_transcription_metrics(
+        instance,
+        duration_histogram,
+        response_dict,
+        duration,
+    )
+
+    # span attributes
+    _set_response_attributes(span, response_dict)
+
+
+def _set_transcription_metrics(
+    instance,
+    duration_histogram,
+    response_dict,
+    duration,
+):
+    from opentelemetry.instrumentation.openai.shared import _get_openai_base_url
+
+    shared_attributes = metric_shared_attributes(
+        response_model=response_dict.get("model") or None,
+        operation="audio.transcriptions",
+        server_address=_get_openai_base_url(instance),
+    )
+
+    # duration metrics
+    if duration and isinstance(duration, (float, int)) and duration_histogram:
+        duration_histogram.record(duration, attributes=shared_attributes)
diff --git a/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/utils.py b/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/utils.py
index 1b16fa3fec..b951f801f6 100644
--- a/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/utils.py
+++ b/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/utils.py
@@ -83,6 +83,29 @@ def wrapper(wrapped, instance, args, kwargs):
     return _with_embeddings_telemetry
 
 
+def _with_audio_telemetry_wrapper(func):
+    """Wrapper to convert the audio wrapper function into the expected format for wrapt."""
+    def _with_audio_telemetry(
+        tracer,
+        duration_histogram,
+        exception_counter,
+    ):
+        def wrapper(wrapped, instance, args, kwargs):
+            return func(
+                tracer,
+                duration_histogram,
+                exception_counter,
+                wrapped,
+                instance,
+                args,
+                kwargs,
+            )
+
+        return wrapper
+
+    return _with_audio_telemetry
+
+
 def _with_chat_telemetry_wrapper(func):
     def _with_chat_telemetry(
         tracer,
diff --git a/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/v1/__init__.py b/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/v1/__init__.py
index 7511a16d6a..e6e9b49bd6 100644
--- a/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/v1/__init__.py
+++ b/packages/opentelemetry-instrumentation-openai/opentelemetry/instrumentation/openai/v1/__init__.py
@@ -18,6 +18,10 @@
 from opentelemetry.instrumentation.openai.shared.image_gen_wrappers import (
     image_gen_metrics_wrapper,
 )
+from opentelemetry.instrumentation.openai.shared.audio_wrappers import (
+    atranscription_wrapper,
+    transcription_wrapper,
+)
 from opentelemetry.instrumentation.openai.utils import is_metrics_enabled
 from opentelemetry.instrumentation.openai.v1.assistant_wrappers import (
     assistants_create_wrapper,
@@ -247,6 +251,36 @@ def _instrument(self, **kwargs):
             image_gen_metrics_wrapper(duration_histogram, image_gen_exception_counter),
         )
 
+        if is_metrics_enabled():
+            audio_transcription_exception_counter = meter.create_counter(
+                # name=Meters.LLM_AUDIO_TRANSCRIPTIONS_EXCEPTIONS, # TODO(Ata): come back here later when semconv is published
+                name='llm.openai.audio.transcriptions.exceptions',
+                unit="time",
+                description="Number of exceptions occurred during audio transcriptions operation",
+            )
+        else:
+            audio_transcription_exception_counter = None
+
+        wrap_function_wrapper(
+            "openai.resources.audio.transcriptions",
+            "Transcriptions.create",
+            transcription_wrapper(
+                tracer,
+                duration_histogram,
+                audio_transcription_exception_counter,
+            ),
+        )
+
+        wrap_function_wrapper(
+            "openai.resources.audio.transcriptions",
+            "AsyncTranscriptions.create",
+            atranscription_wrapper(
+                tracer,
+                duration_histogram,
+                audio_transcription_exception_counter,
+            ),
+        )
+
         # Beta APIs may not be available consistently in all versions
         self._try_wrap(
             "openai.resources.beta.assistants",
@@ -338,6 +372,8 @@ def _uninstrument(self, **kwargs):
         unwrap("openai.resources.completions", "AsyncCompletions.create")
         unwrap("openai.resources.embeddings", "AsyncEmbeddings.create")
         unwrap("openai.resources.images", "Images.generate")
+        unwrap("openai.resources.audio.transcriptions", "Transcriptions.create")
+        unwrap("openai.resources.audio.transcriptions", "AsyncTranscriptions.create")
 
         # Beta APIs may not be available consistently in all versions
         try:
diff --git a/packages/opentelemetry-instrumentation-openai/pyproject.toml b/packages/opentelemetry-instrumentation-openai/pyproject.toml
index 3abf68f228..0a3076d05b 100644
--- a/packages/opentelemetry-instrumentation-openai/pyproject.toml
+++ b/packages/opentelemetry-instrumentation-openai/pyproject.toml
@@ -28,6 +28,7 @@ opentelemetry-api = "^1.38.0"
 opentelemetry-instrumentation = ">=0.59b0"
 opentelemetry-semantic-conventions = ">=0.59b0"
 opentelemetry-semantic-conventions-ai = "^0.4.13"
+mutagen = { version = "^1.47.0", optional = true }
 
 [tool.poetry.group.dev.dependencies]
 autopep8 = "^2.2.0"
@@ -49,6 +50,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry.extras]
 instruments = ["openai"]
+audio = ["mutagen"]
 
 [tool.poetry.plugins."opentelemetry_instrumentor"]
 openai = "opentelemetry.instrumentation.openai:OpenAIInstrumentor"
diff --git a/packages/opentelemetry-instrumentation-openai/tests/traces/test_audio_transcription.py b/packages/opentelemetry-instrumentation-openai/tests/traces/test_audio_transcription.py
new file mode 100644
index 0000000000..bacd82a2bb
--- /dev/null
+++ b/packages/opentelemetry-instrumentation-openai/tests/traces/test_audio_transcription.py
@@ -0,0 +1,60 @@
+import io
+import pytest
+from opentelemetry.semconv._incubating.attributes import (
+    gen_ai_attributes as GenAIAttributes,
+)
+from opentelemetry.semconv_ai import SpanAttributes
+
+
+@pytest.mark.vcr
+def test_audio_transcription(instrument_legacy, span_exporter, openai_client):
+    # Create a mock audio file (in real test, use VCR cassette)
+    audio_file = io.BytesIO(b"fake audio content")
+    audio_file.name = "test_audio.mp3"
+
+    openai_client.audio.transcriptions.create(
+        file=audio_file,
+        model="whisper-1",
+    )
+
+    spans = span_exporter.get_finished_spans()
+    assert len(spans) == 1
+    assert spans[0].name == "openai.audio.transcriptions"
+
+    transcription_span = spans[0]
+    assert (
+        transcription_span.attributes[GenAIAttributes.GEN_AI_REQUEST_MODEL]
+        == "whisper-1"
+    )
+    assert (
+        transcription_span.attributes[SpanAttributes.LLM_OPENAI_API_BASE]
+        == "https://api.openai.com/v1/"
+    )
+
+
+@pytest.mark.vcr
+async def test_audio_transcription_async(
+    instrument_legacy, span_exporter, async_openai_client
+):
+    # Create a mock audio file
+    audio_file = io.BytesIO(b"fake audio content")
+    audio_file.name = "test_audio.mp3"
+
+    await async_openai_client.audio.transcriptions.create(
+        file=audio_file,
+        model="whisper-1",
+    )
+
+    spans = span_exporter.get_finished_spans()
+    assert len(spans) == 1
+    assert spans[0].name == "openai.audio.transcriptions"
+
+    transcription_span = spans[0]
+    assert (
+        transcription_span.attributes[GenAIAttributes.GEN_AI_REQUEST_MODEL]
+        == "whisper-1"
+    )
+    assert (
+        transcription_span.attributes[SpanAttributes.LLM_OPENAI_API_BASE]
+        == "https://api.openai.com/v1/"
+    )