Upload kalbos_nustatymas.py

kalbos_nustatymas.py

@@ -0,0 +1,53 @@
+
+import whisper
+from transformers import pipeline
+import torch
+import torchaudio
+
+# 🔹 Whisper transkripcija
+def transcribe_text(audio_path):
+    model = whisper.load_model("base")
+    result = model.transcribe(audio_path, language='lt')
+    return result.get("text", "").strip()
+
+# 🔹 Whisper kalbos atpažinimas (su papildomu raktinių žodžių tikrinimu)
+def recognize_language(audio_path):
+    model = whisper.load_model("base")
+    result = model.transcribe(audio_path)
+    text = result.get("text", "").strip()
+    lang_code = result.get("language", "unknown")
+
+    lower_text = text.lower()
+    if any(word in lower_text for word in ["labas", "ačiū", "draugas", "vardas", "sekasi", "prašau"]):
+        return "lt"
+    elif any(word in lower_text for word in ["hello", "name", "how are you", "friend", "please"]):
+        return "en"
+    elif any(word in lower_text for word in ["hallo", "danke", "freund", "ich", "bitte"]):
+        return "de"
+    else:
+        return lang_code
+
+# 🔸 Wav2Vec2 transkripcija (su kalbos pasirinkimu)
+def transcribe_text_wav2vec(audio_path, kalba):
+    kalbos_modeliai = {
+        "lt": "DeividasM/wav2vec2-large-xlsr-53-lithuanian",
+        "en": "facebook/wav2vec2-base-960h",
+        "de": "jonatasgrosman/wav2vec2-large-xlsr-53-german"
+    }
+
+    if kalba not in kalbos_modeliai:
+        raise ValueError(f"Nepalaikoma kalba: {kalba}")
+
+    pipe = pipeline(
+        "automatic-speech-recognition",
+        model=kalbos_modeliai[kalba]
+    )
+
+    speech_array, sampling_rate = torchaudio.load(audio_path)
+    if sampling_rate != 16000:
+        resampler = torchaudio.transforms.Resample(orig_freq=sampling_rate, new_freq=16000)
+        speech_array = resampler(speech_array)
+    speech = speech_array[0].numpy()
+
+    result = pipe(speech)
+    return result["text"]