Support speaker names in voice assignment

Allow voices to be assigned using speaker names/roles in addition
to speaker order (e.g., speaker_1, speaker_2).

Speaker names are case-insensitive, so "Doctor", "DOCTOR",
and "doctor" are treated equivalently.

Examples:

```python
# Assigning voices using speaker names / roles
dialog.to_audio(
    "path/to/output/audio",
    voices={
        "Doctor": "path/to/doctor_voice_reference.wav",
        "Patient": "path/to/patient_voice_reference.wav",
    }
)

# Assigning voices using speaker orders (speaker_1, speaker_2)
dialog.to_audio(
    "path/to/output/audio",
    voices={
        "speaker_1": "path/to/doctor_voice_reference.wav",
        "speaker_2": "path/to/patient_voice_reference.wav",
    }
)
```

Speaker names are case insensitive, so this will also work exactly the same as the first example:
```python
dialog.to_audio(
    "path/to/output/audio",
    voices={
        "DOCTOR": "path/to/doctor_voice_reference.wav",
        "PATIENT": "path/to/patient_voice_reference.wav",
    }
)
```

This keeps backward compatibility while making voice
configuration more flexible and readable.

Author: sergioburdisso

src/sdialog/audio/dialog.py

@@ -46,9 +46,9 @@
 from sdialog import Dialog
 from sdialog.audio.turn import AudioTurn
 from sdialog.audio.room import AudioSource
-from sdialog.audio.utils import logger, generate_reference_voices, Role
 from sdialog.audio.tts.base import BaseTTS, BaseVoiceCloneTTS
 from sdialog.audio.voice_database import BaseVoiceDatabase, Voice
+from sdialog.audio.utils import Role, CaseInsensitiveDict, logger, generate_reference_voices
 
 
 class AudioDialog(Dialog):
@@ -68,8 +68,8 @@ class AudioDialog(Dialog):
     audio_step_2_filepath: str = ""
     audio_step_3_filepaths: dict[str, dict] = {}
 
-    speakers_names: dict[str, str] = {}
-    speakers_roles: dict[str, str] = {}
+    speakers_names: dict[str, str] = {}  # role2name mapping
+    speakers_roles: dict[str, str] = CaseInsensitiveDict()  # name2role mapping (case insensitive)
 
     def __init__(self, **kwargs):
         super().__init__(**kwargs)
@@ -177,16 +177,14 @@ def from_dialog(dialog: Dialog):
                 speakers.append(turn.speaker)
                 if len(speakers) == 2:
                     break
-        speaker_1 = speakers[0]
-        speaker_2 = speakers[1]
 
         # Create role mappings for speaker identification
-        audio_dialog.speakers_names[Role.SPEAKER_1] = speaker_1
-        audio_dialog.speakers_names[Role.SPEAKER_2] = speaker_2
+        audio_dialog.speakers_names[Role.SPEAKER_1] = speakers[0]
+        audio_dialog.speakers_names[Role.SPEAKER_2] = speakers[1]
 
         # Create reverse mappings for role lookup
-        audio_dialog.speakers_roles[speaker_1] = Role.SPEAKER_1
-        audio_dialog.speakers_roles[speaker_2] = Role.SPEAKER_2
+        audio_dialog.speakers_roles[speakers[0]] = Role.SPEAKER_1
+        audio_dialog.speakers_roles[speakers[1]] = Role.SPEAKER_2
 
         return audio_dialog
 
@@ -478,8 +476,8 @@ def persona_to_voice(
             # Get the role of the speaker (speaker_1 or speaker_2)
             role: Role = self.speakers_roles[speaker]
 
-            if voices is not None and voices != {} and role not in voices:
-                raise ValueError(f"Voice for role {str(role)} not found in the voices dictionary")
+            if voices and role not in voices and speaker not in voices:
+                raise ValueError(f"Voice for {str(role)} not found in the voices dictionary")
 
             # If no voices are provided, get a voice from the voice database based on the gender, age and language
             if voices is None or voices == {}:

src/sdialog/audio/pipeline.py

@@ -543,6 +543,10 @@ def inference(
         if not verbose:
             logger.setLevel(logging.ERROR)
 
+        if voices is not None:
+            voices = {dialog.speakers_roles[key] if key in dialog.speakers_roles else key: value
+                      for key, value in voices.items()}
+
         # Reset the logger level to the original level after the function is executed
         try:
 

src/sdialog/audio/utils.py

@@ -424,6 +424,30 @@ def __str__(self):
         return self.value
 
 
+class CaseInsensitiveDict(dict):
+    def __setitem__(self, key, value):
+        super().__setitem__(key.lower(), value)
+
+    def __getitem__(self, key):
+        return super().__getitem__(key.lower())
+
+    def __delitem__(self, key):
+        super().__delitem__(key.lower())
+
+    def __contains__(self, key):
+        return super().__contains__(key.lower())
+
+    def get(self, key, default=None):
+        return super().get(key.lower(), default)
+
+    def update(self, other=None, **kwargs):
+        if other:
+            for k, v in dict(other).items():
+                self[k] = v
+        for k, v in kwargs.items():
+            self[k] = v
+
+
 def default_dscaper_datasets() -> list[str]:
     """
     Default dSCAPER datasets

tests/test_audio.py

@@ -693,7 +693,7 @@ def test_persona_to_voice_missing_role_in_voices_dict(dialog_with_personas):
         # SPEAKER_2 is missing
     }
 
-    with pytest.raises(ValueError, match="Voice for role speaker_2 not found in the voices dictionary"):
+    with pytest.raises(ValueError):
         dialog_with_personas.persona_to_voice(mock_voice_db, voices=voices)