Speech Emotion Recognition Model trained on litagin/Galgame_Speech_SER_16kHz.
Usage
from transformers import pipeline
REPO_ID = "litagin/anime_speech_emotion_classification"
pipe = pipeline(
"audio-classification",
model=REPO_ID,
feature_extractor=REPO_ID,
trust_remote_code=True,
device="cuda",
)
audio_path = "path/to/audio.wav"
result = pipe(audio_path)
print(result)
Result:
[{'score': 0.5655683279037476, 'label': 'Angry'},
{'score': 0.12489483505487442, 'label': 'Disgusted'},
{'score': 0.11449059844017029, 'label': 'Embarrassed'},
{'score': 0.06627542525529861, 'label': 'Surprised'},
{'score': 0.06157735362648964, 'label': 'Sad'},
{'score': 0.031055787578225136, 'label': 'Neutral'},
{'score': 0.022820966318249702, 'label': 'Happy'},
{'score': 0.00791135337203741, 'label': 'Fearful'},
{'score': 0.00540440296754241, 'label': 'Sexual1'},
{'score': 8.61035857724346e-07, 'label': 'Sexual2'}]
Label
"id2label": {
"0": "Angry",
"1": "Disgusted",
"2": "Embarrassed",
"3": "Fearful",
"4": "Happy",
"5": "Sad",
"6": "Surprised",
"7": "Neutral",
"8": "Sexual1", # NSFW erotic voices such as 喘ぎ
"9": "Sexual2" # Blowjob Oral Slurp SFX チュパ音
},