Spaces:
Sleeping
Sleeping
| import os, requests, shutil | |
| from pydub import AudioSegment | |
| import whisper | |
| from speechbrain.pretrained.interfaces import foreign_class | |
| class AccentAnalyzerTool: | |
| def __init__(self): | |
| self.whisper_model = whisper.load_model("medium") | |
| self.accent_model = foreign_class( | |
| source="Jzuluaga/accent-id-commonaccent_xlsr-en-english", | |
| pymodule_file="custom_interface.py", | |
| classname="CustomEncoderWav2vec2Classifier" | |
| ) | |
| self.last_transcript = None | |
| def log(self, msg): | |
| print(f"[AccentAnalyzerTool] {msg}") | |
| def analyze(self, url: str) -> str: | |
| try: | |
| self.log("Downloading video...") | |
| tmp_dir = "tmp" | |
| os.makedirs(tmp_dir, exist_ok=True) | |
| video_path = os.path.join(tmp_dir, "video.mp4") | |
| r = requests.get(url) | |
| with open(video_path, "wb") as f: | |
| f.write(r.content) | |
| self.log("Extracting audio...") | |
| audio_path = os.path.join(tmp_dir, "audio.wav") | |
| AudioSegment.from_file(video_path).export(audio_path, format="wav") | |
| self.log("Classifying accent...") | |
| _, score, _, label = self.accent_model.classify_file(audio_path) | |
| accent = label[0].upper() if label[0] == 'us' else label[0].capitalize() | |
| confidence = round(float(score) * 100, 2) | |
| self.log("Transcribing...") | |
| transcript = self.whisper_model.transcribe(audio_path)["text"] | |
| self.last_transcript = transcript | |
| summary = ( | |
| f"The speaker has a **{accent} English accent** " | |
| f"with **{confidence}% confidence**.\n\n" | |
| f"**Transcript of the audio:**\n\n *{transcript.strip(' ')}*" | |
| ) | |
| shutil.rmtree(tmp_dir, ignore_errors=True) | |
| return summary | |
| except Exception as e: | |
| return f"Error analyzing accent: {str(e)}" | |