Add language detection test for Hindi audio transcription (#356)

* Add language detection test for Hindi audio transcription

- Added test_language_detection_hindi() to verify language detection functionality
- Uses Hindi audio file OSR_in_000_0062_16k.wav from voiptroubleshooter.com
- Validates that whisper-large-v3 correctly identifies Hindi language as "hi"
- Test uses verbose_json format to access language metadata

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* replace url

---------

Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Sahil Yadav <yadavsahil197@gmail.com>

Author: 3 people

tests/integration/resources/test_transcriptions.py

@@ -22,9 +22,7 @@ def test_basic_transcription_url(self, sync_together_client):
         """
         Test basic transcription with URL audio file
         """
-        audio_url = (
-            "https://voiptroubleshooter.com/open_speech/american/OSR_us_000_0010_8k.wav"
-        )
+        audio_url = "https://together-public-test-data.s3.us-west-2.amazonaws.com/audio/test_5s_clip.wav"
 
         response = sync_together_client.audio.transcriptions.create(
             file=audio_url, model="openai/whisper-large-v3"
@@ -38,9 +36,7 @@ def test_transcription_with_language(self, sync_together_client):
         """
         Test transcription with language parameter
         """
-        audio_url = (
-            "https://voiptroubleshooter.com/open_speech/american/OSR_us_000_0010_8k.wav"
-        )
+        audio_url = "https://together-public-test-data.s3.us-west-2.amazonaws.com/audio/test_5s_clip.wav"
 
         response = sync_together_client.audio.transcriptions.create(
             file=audio_url, model="openai/whisper-large-v3", language="en"
@@ -54,9 +50,7 @@ def test_transcription_verbose_json(self, sync_together_client):
         """
         Test transcription with verbose JSON format and timestamps
         """
-        audio_url = (
-            "https://voiptroubleshooter.com/open_speech/american/OSR_us_000_0010_8k.wav"
-        )
+        audio_url = "https://together-public-test-data.s3.us-west-2.amazonaws.com/audio/test_5s_clip.wav"
 
         response = sync_together_client.audio.transcriptions.create(
             file=audio_url,
@@ -74,9 +68,7 @@ def test_transcription_with_temperature(self, sync_together_client):
         """
         Test transcription with temperature parameter
         """
-        audio_url = (
-            "https://voiptroubleshooter.com/open_speech/american/OSR_us_000_0010_8k.wav"
-        )
+        audio_url = "https://together-public-test-data.s3.us-west-2.amazonaws.com/audio/test_5s_clip.wav"
 
         response = sync_together_client.audio.transcriptions.create(
             file=audio_url, model="openai/whisper-large-v3", temperature=0.2
@@ -99,12 +91,28 @@ def test_transcription_missing_model(self, sync_together_client):
         """
         Test transcription with missing model parameter - should use default model
         """
-        audio_url = (
-            "https://voiptroubleshooter.com/open_speech/american/OSR_us_000_0010_8k.wav"
-        )
+        audio_url = "https://together-public-test-data.s3.us-west-2.amazonaws.com/audio/test_5s_clip.wav"
 
         response = sync_together_client.audio.transcriptions.create(file=audio_url)
 
         assert isinstance(response, AudioTranscriptionResponse)
         assert isinstance(response.text, str)
         assert len(response.text) > 0
+
+    def test_language_detection_hindi(self, sync_together_client):
+        """
+        Test language detection with Hindi audio file
+        """
+        audio_url = "https://together-public-test-data.s3.us-west-2.amazonaws.com/audio/hindi_audio.wav"
+
+        response = sync_together_client.audio.transcriptions.create(
+            file=audio_url,
+            model="openai/whisper-large-v3",
+            response_format="verbose_json",
+        )
+
+        assert isinstance(response, AudioTranscriptionVerboseResponse)
+        assert isinstance(response.text, str)
+        assert len(response.text) > 0
+        assert hasattr(response, "language")
+        assert response.language == "hi"