app.py

import os
from math import floor
from typing import Optional

import spaces
import torch
import gradio as gr
from transformers import pipeline
from transformers.pipelines.audio_utils import ffmpeg_read

# config
model_name = "kotoba-tech/kotoba-whisper-v2.2"
example_file = "sample_diarization_japanese.mp3"
if torch.cuda.is_available():
    pipe = pipeline(
        model=model_name,
        chunk_length_s=15,
        batch_size=16,
        torch_dtype=torch.bfloat16,
        device="cuda",
        model_kwargs={'attn_implementation': 'sdpa'},
        trust_remote_code=True
    )
else:
    pipe = pipeline(model=model_name, chunk_length_s=15, batch_size=16, trust_remote_code=True)


def format_time(start: Optional[float], end: Optional[float]):

    def _format_time(seconds: Optional[float]):
        if seconds is None:
            return "[no timestamp available]"
        minutes = floor(seconds / 60)
        hours = floor(seconds / 3600)
        seconds = seconds - hours * 3600 - minutes * 60
        m_seconds = floor(round(seconds - floor(seconds), 1) * 10)
        seconds = floor(seconds)
        return f'{minutes:02}:{seconds:02}.{m_seconds:01}'

    return f"[{_format_time(start)} -> {_format_time(end)}]:"


@spaces.GPU
def get_prediction(inputs, **kwargs):
    return pipe(inputs, **kwargs)


def transcribe(inputs: str,
               add_punctuation: bool,
               num_speakers: float,
               min_speakers: float,
               max_speakers: float,
               chunk_length_s: float,
               add_silence_end: float,
               add_silence_start: float):
    if inputs is None:
        raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
    with open(inputs, "rb") as f:
        inputs = f.read()
    array = ffmpeg_read(inputs, pipe.feature_extractor.sampling_rate)
    prediction = get_prediction(
        inputs={"array": array, "sampling_rate": pipe.feature_extractor.sampling_rate},
        add_punctuation=add_punctuation,
        num_speakers=int(num_speakers) if num_speakers != 0 else None,
        min_speakers=int(min_speakers) if min_speakers != 0 else None,
        max_speakers=int(max_speakers) if max_speakers != 0 else None,
        chunk_length_s=int(chunk_length_s) if chunk_length_s != 30 else None,
        add_silence_end=add_silence_end if add_silence_end != 0 else None,
        add_silence_start=add_silence_start if add_silence_start != 0 else None
    )
    output = ""
    for n, s in enumerate(prediction["speaker_ids"]):
        text_timestamped = "\n".join([f"- **{format_time(*c['timestamp'])}** {c['text']}" for c in prediction[f"chunks/{s}"]])
        output += f'### Speaker {n+1} \n{prediction[f"text/{s}"]}\n\n{text_timestamped}\n'
    return output


description = (f"Transcribe and diarize long-form microphone or audio inputs with the click of a button! Demo uses "
               f"Kotoba-Whisper [{model_name}](https://huggingface.co/{model_name}).")
title = f"Audio Transcription and Diarization with {os.path.basename(model_name)}"
shared_config = {"fn": transcribe, "title": title, "description": description, "allow_flagging": "never", "examples": [
    [example_file, True, 0, 0, 0, 30, 0.5, 0.5],
    [example_file, True, 4, 0, 0, 30, 0.5, 0.5]
]}
o_upload = gr.Markdown()
o_mic = gr.Markdown()
options = [

]
i_upload = gr.Interface(
    inputs=[
        gr.Audio(sources="upload", type="filepath", label="Audio file"),
        gr.Checkbox(label="add punctuation", value=True),
        gr.Slider(0, 10, label="num speakers (set 0 for auto-detect mode)", value=0, step=1),
        gr.Slider(0, 10, label="min speakers (set 0 for auto-detect mode)", value=0, step=1),
        gr.Slider(0, 10, label="max speakers (set 0 for auto-detect mode)", value=0, step=1),
        gr.Slider(5, 30, label="chunk length for ASR", value=30, step=1),
        gr.Slider(0, 0.5, label="silence at the end", value=0.5, step=0.05),
        gr.Slider(0, 0.5, label="silence at the start", value=0.5, step=0.05),
    ],
    outputs=gr.Markdown(),
    **shared_config
)
i_mic = gr.Interface(
    inputs=[
        gr.Audio(sources="microphone", type="filepath", label="Microphone input"),
        gr.Checkbox(label="add punctuation", value=True),
        gr.Slider(0, 10, label="num speakers (set 0 for auto-detect mode)", value=0, step=1),
        gr.Slider(0, 10, label="min speakers (set 0 for auto-detect mode)", value=0, step=1),
        gr.Slider(0, 10, label="max speakers (set 0 for auto-detect mode)", value=0, step=1),
        gr.Slider(5, 30, label="chunk length for ASR", value=30, step=1),
        gr.Slider(0, 0.5, label="silence at the end", value=0.5, step=0.05),
        gr.Slider(0, 0.5, label="silence at the start", value=0.5, step=0.05),
    ],
    outputs=gr.Markdown(),
    **shared_config
)
with gr.Blocks() as demo:
    gr.TabbedInterface([i_upload, i_mic], ["Audio file", "Microphone"])
demo.queue(api_open=False, default_concurrency_limit=40).launch(show_api=False, show_error=True)