Whisper Large Turbo in Free Google Colab for Transcription - Step-by-Step Tutorial

 This video shows how to install and use Whisper large-v3-turbo in free google colab for transcription in gradio which is a finetuned version of a pruned Whisper large-v3. 

Code:

!pip install git+https://github.com/huggingface/transformers gradio

import torch

from transformers import pipeline

pipe = pipeline("automatic-speech-recognition",

               "openai/whisper-large-v3-turbo",

               torch_dtype=torch.float16,

               device="cuda:0")

pipe("/content/samples_jfk.wav")

import gradio as gr

def transcribe(inputs):

    if inputs is None:

        raise gr.Error("No audio file")

    text = pipe(inputs, generate_kwargs={"task": "transcribe"}, return_timestamps=True)["text"]

    return text

demo = gr.Interface(

    fn=transcribe,

    inputs=[

        gr.Audio(sources=["microphone", "upload"], type="filepath"),

    ],

    outputs="text",

    title="Whisper Large V3 Turbo: Transcribe Audio",

    description=(

        "Transcribe long-form microphone or audio inputs. Thanks to HuggingFace"

    ),

    allow_flagging="never",

)

demo.launch()