feat:finetuned models

allow fine tuned models to be used for each language

examples:
- high accuracy for target lang https://huggingface.co/projecte-aina/faster-whisper-large-v3-ca-3catparla
- lots of tiny models for each language https://huggingface.co/ammaraldirawi

README.md

@@ -9,13 +9,17 @@ High-performance inference of [OpenAI's Whisper](https://github.com/openai/whisp
 
 `pip install ovos-stt-plugin-fasterwhisper`
 
-## Configuration
+## Models
+
+available models are `'tiny.en', 'tiny', 'base.en', 'base', 'small.en', 'small', 'medium.en', 'medium', 'large-v1', 'large-v2', 'large-v3', 'large', 'distil-large-v2', 'distil-medium.en', 'distil-small.en', 'distil-large-v3'`
 
-available models are `"tiny.en", "tiny", "base.en", "base", "small.en", "small", "medium.en", "medium", "large-v2", "large-v3"`
+you can also pass a full path to a local model or a huggingface repo_id, eg. `"projecte-aina/whisper-large-v3-ca-3catparla"`
 
-eg, to use Large model with GPU
+You can [convert](https://github.com/SYSTRAN/faster-whisper?tab=readme-ov-file#model-conversion) any whisper model, or use any [compatible model from huggingface](https://huggingface.co/models?pipeline_tag=automatic-speech-recognition&sort=modified&search=faster-whisper)
+
+## Configuration
 
-To use Whisper as STT
+to use Large model with GPU
 
 ```json
   "stt": {

ovos_stt_plugin_fasterwhisper/__init__.py

@@ -173,12 +173,11 @@ class FasterWhisperSTT(STT):
 
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)
-        model = self.config.get("model")
+        model = self.config.get("model") or "small"
         valid_model = model in FasterWhisperSTT.MODELS
-        if not model or not valid_model:
-            LOG.warning(f"{model} is not a valid model ({FasterWhisperSTT.MODELS}), using 'small' instead")
-            model = "small"
-            self.config["model"] = "small"
+        if not valid_model:
+            LOG.info(f"{model} is not default model_id ({FasterWhisperSTT.MODELS}), "
+                     f"assuming huggingface repo_id or path to local model file")
 
         self.beam_size = self.config.get("beam_size", 5)
         self.compute_type = self.config.get("compute_type", "int8")