HaHaScore

🎬 Try it on a YouTube Video

Paste any YouTube URL and our AI will analyze the funniness across the whole video.

🤖 How It Works

HaHaScore Cascade v10 uses a multimodal AI architecture that fuses text + audio cues:

💻 Use the Model in Your Code

import onnxruntime as ort
import numpy as np
from huggingface_hub import hf_hub_download

# Load model
model_path = hf_hub_download(
    repo_id="Hayasuki/hahascore-cascade-v10",
    filename="v10_cascade_int8.onnx"
)
session = ort.InferenceSession(model_path)

# Run inference
text = np.random.randn(1, 20, 768).astype(np.float32)
audio = np.random.randn(1, 20, 791).astype(np.float32)
cross = np.zeros((1, 20, 16), dtype=np.float32)

scores, conf, gate = session.run(None, {
    'text': text, 'audio': audio, 'cross': cross
})
print(f"Average humor: {scores.mean():.3f}")