| import gradio as gr | |
| import numpy as np | |
| import onnx_asr | |
| models = {name: onnx_asr.load_model(name) for name in ["alphacep/vosk-model-ru", "alphacep/vosk-model-small-ru"]} | |
| def recoginize(audio: tuple[int, np.ndarray]): | |
| sample_rate, waveform = audio | |
| waveform = waveform.astype(np.float32) / 2 ** (8 * waveform.itemsize - 1) | |
| return [[name, model.recognize(waveform, sample_rate=sample_rate)] for name, model in models.items()] | |
| demo = gr.Interface( | |
| fn=recoginize, | |
| inputs=[gr.Audio(min_length=1, max_length=10)], | |
| outputs=[gr.Dataframe(headers=["Model", "result"], wrap=True, show_fullscreen_button=True)], | |
| flagging_mode="never", | |
| ) | |
| demo.launch() | |