Nothing Special   »   [go: up one dir, main page]
Scribd Logo
Upload

Welcome to Scribd!

  • 51 views

    Ai ML Assignment 2 - Voice Cloning

    Uploaded by

    Abu Thalib
    The document provides details about an assignment to build a voice cloning model. Specifically, it asks the applicant to: 1) Create a model that can generate synthetic speech that matches the vocal characteristics of a target speaker based on text input. 2) The model should be able to accurately reproduce the target speaker's accent, pitch, speed and generate clear, understandable speech. 3) The model code and a sample output are to be submitted on GitHub along with a README file explaining how to use the model within 24-48 hours. Using a pre-trained model is recommended over training a new model.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Ai ML Assignment 2 - Voice Cloning

    Uploaded by

    Abu Thalib
    51 views4 pages
    The document provides details about an assignment to build a voice cloning model. Specifically, it asks the applicant to: 1) Create a model that can generate synthetic speech that matches the vocal characteristics of a target speaker based on text input. 2) The model should be able to accurately reproduce the target speaker's accent, pitch, speed and generate clear, understandable speech. 3) The model code and a sample output are to be submitted on GitHub along with a README file explaining how to use the model within 24-48 hours. Using a pre-trained model is recommended over training a new model.

    Original Description:

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document provides details about an assignment to build a voice cloning model. Specifically, it asks the applicant to: 1) Create a model that can generate synthetic speech that matches the vocal characteristics of a target speaker based on text input. 2) The model should be able to accurately reproduce the target speaker's accent, pitch, speed and generate clear, understandable speech. 3) The model code and a sample output are to be submitted on GitHub along with a README file explaining how to use the model within 24-48 hours. Using a pre-trained model is recommended over training a new model.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    51 views4 pages

    Ai ML Assignment 2 - Voice Cloning

    Uploaded by

    Abu Thalib
    The document provides details about an assignment to build a voice cloning model. Specifically, it asks the applicant to: 1) Create a model that can generate synthetic speech that matches the vocal characteristics of a target speaker based on text input. 2) The model should be able to accurately reproduce the target speaker's accent, pitch, speed and generate clear, understandable speech. 3) The model code and a sample output are to be submitted on GitHub along with a README file explaining how to use the model within 24-48 hours. Using a pre-trained model is recommended over training a new model.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 4

    Assignment - Ai/Ml

    About Listed- https://listed.fans/


    We are a lab building products for the creator economy. Why so, because we are on a mission to
    empower the individual to realize/unleash their full potential. Following are some of our flagship
    products:

    OpeninApp
    Almost all your favorite creators on Instagram share links to their latest YouTube videos or YouTube
    channels in their bio or stories. But whenever you tap on those links, you get redirected to the YouTube
    website instead of the YouTube app. And here you can’t watch the video in full screen nor can you like it,
    comment on it or subscribe to the channel.
    Ever wondered why it is so? That’s because Instagram has something called a Built—IN-Browser that
    opens up instead of opening the app you wanted to open. Which is not cool at all….And this is not just
    so for IG but for all your favorite social media apps eg. YouTube, Linkedin, Twitter, Snapchat, Telegram,
    Quora etc.
    So we fixed this problem of sub-optimal user experiences by building a tool that ensures all social media
    links open in the apps they should. It’s called OpeninApp. And it is already loved by over 20,000
    influencers & brands including the likes of Munawar Faruqui, Raj Shamani, Riva Arora, Aishwarya
    (Ranjnikant-ji’s daughter), Trakin Tech, Mamaearth etc. And has been used by 55Mn+ users.

    Listed.fans
    We Listians find it very painful that unlike tech entrepreneurs there’s little to no capital available for our
    darling social media creators or influencers, the wonderful people who entertain us, educate us and so so
    much more. And this is not just the case for social media creators but individual performers of any form
    are they are, cricketers, singers, dancers, actors etc… So, Listed helps rising creators get funded, just like
    startups by being there done-that's like mega-influencers, agency owners and brand CXOs to launch and
    accelerate their own creator-ships. In return, these rising stars pledge to share a portion of their future
    incomes with their investors for a certain fixed period of time.
    So, if you believe the gifted individuals of today will overshadow the companies of yesterday, and in
    doing so, have the right to capital & guidance and you are passionate about social media and wants to
    improve User Experiences for a Billion people, then welcome home.
    TASK

    Create a voice cloning model that can generate a synthetic voice that sounds like a specific person. The model
    should be able to generate speech from text input, and it should be able to reproduce the unique vocal
    characteristics of the target speaker.

    What are we looking for?

    ● The model must be able to generate speech that is indistinguishable from the target speaker's natural
    voice.
    ● The model must be able to copy the accent of the original audio.
    ● The model must be able to generate speech at a variety of speeds and pitches.
    ● The model must be able to generate speech that is clear and understandable.

    Your assignment will be evaluated on the following criteria -

    ● Accuracy: How closely does the generated speech resemble the target speaker's natural voice?
    ● Fluency: How natural and smooth is the generated speech?
    ● Understandability: How clear and easy to understand is the generated speech?
    ● Code quality, the less the code the better.
    ● Time taken to submit, the less the better.
    ● Response times, the faster the better.

    Additional Requirements:

    ● The model must be implemented in Python.


    ● The model must be accompanied by a README file that explains how to run the model and how to evaluate
    its performance.
    Submission:
    ● The model must be submitted as a GitHub repository. The repository must contain the following:
    ● The Python code for the model.
    ● The README file.
    ● A Sample Output along with the inputs given

    By When can you complete the task:

    24 to 48 hours - this is an urgent opening and if you have received the assignment you are already ahead
    of many candidates, hence the faster you submit clean readable code along with the video demo the
    faster we can move ahead with your application.

    Feel free to write, if you have any doubts. Please follow these given instructions while
    dropping mails -

    1. Write the subject in the given format - (Your Name) - (Exact Name of your
    Position) - (Purpose of Mail).
    2. Write to muskaan@openinapp.com / careers@listed.fans

    FAQs:

    1. How can I submit the assignment?

    Ans - Submit the Video on the same mail from which you have received
    2. Can I use a pre-trained model, or do I need to train a model myself?

    Ans - Using pre-trained models is usually more practical because training models can take a long time.
    We kindly request that you prioritize using the provided link to access the pre-trained model.

    You might also like