---
configs:
- config_name: default
  data_files:
  - split: Anthony_Blinken
    path: data/Anthony_Blinken-*
  - split: Barack_Obama
    path: data/Barack_Obama-*
  - split: Donald_Trump
    path: data/Donald_Trump-*
  - split: Elon_Musk
    path: data/Elon_Musk-*
  - split: JD_Vance
    path: data/JD_Vance-*
  - split: Joe_Biden
    path: data/Joe_Biden-*
  - split: Kamala_Harris
    path: data/Kamala_Harris-*
  - split: Mathew_Miller
    path: data/Mathew_Miller-*
  - split: Tim_Walz
    path: data/Tim_Walz-*
  - split: Vivek_Ramaswamy
    path: data/Vivek_Ramaswamy-*
dataset_info:
  features:
  - name: Audio
    dtype: audio
  - name: Label
    dtype: string
  - name: Source
    dtype: string
  - name: Speaker
    dtype: string
  - name: AudioName
    dtype: string
  splits:
  - name: Anthony_Blinken
    num_bytes: 9288074593
    num_examples: 19403
  - name: Barack_Obama
    num_bytes: 91170450928
    num_examples: 167793
  - name: Donald_Trump
    num_bytes: 4719985579
    num_examples: 11320
  - name: Elon_Musk
    num_bytes: 10322815043.88
    num_examples: 15012
  - name: JD_Vance
    num_bytes: 4526683330.6
    num_examples: 7556
  - name: Joe_Biden
    num_bytes: 40413289908.84
    num_examples: 48034
  - name: Kamala_Harris
    num_bytes: 32463183347.4
    num_examples: 35030
  - name: Mathew_Miller
    num_bytes: 12861086794.84
    num_examples: 16578
  - name: Tim_Walz
    num_bytes: 1774198427.016
    num_examples: 4968
  - name: Vivek_Ramaswamy
    num_bytes: 21034647753.44
    num_examples: 22442
  download_size: 70922268576
  dataset_size: 123395904606.016
language:
- en
tags:
- sasv
- asv
- tts
- speaker-specific
- source-tracing
- anti-spoofing
- deepfake
pretty_name: FamousFigures
extra_gated_heading: Acknowledge license to accept the repository
extra_gated_description: >-
  Our team may take 2-3 days to process your request. Please use an email
  address that is from your current affiliation, not personal email addresses
  (e.g., gmail). To avoid anonymous downloads, we accept officially affiliated
  email addresses.
extra_gated_button_content: Acknowledge license
extra_gated_prompt: >-
  By accepting and downloading the dataset, you agree to the license and
  restrictions. Please check https://datasets.issflab.net/Famous_fig.html for
  full details. We may send important updates or license-related information to
  the registered e-mail address.
extra_gated_fields:
  Full name: text
  E-mail: text
  Affiliation: text
  Country: country
  Date of agreement: date_picker
  You are from:
    type: select
    options:
    - Research
    - Industry
    - Both
license: cc-by-4.0
---
<!-- Drop this block at the top of your index.html or README.md -->
<h1>Famous Figures: Collecting, Curating, and Annotating Good Quality Speech Deepfake Dataset for Famous Figures — Process and Challenges</h1>

<p align="center">
  <a href="https://arxiv.org/abs/2507.00324">Paper</a>,
  <a href="https://datasets.issflab.net/Famous_fig.html">Project Page</a>
</p>
<h3>Authors</h3>
Hashim Ali, Surya Subramani, Raksha Varahamurthy, Nithin Adupa, Lekha Bollinani, Hafiz Malik


<!-- <h3>Dataset Creators</h3>
Surya Subramani, Hashim Ali, Raksha Varahamurthy -->
<p align="center">
  <i>Interspeech 2025</i>
</p>
<h3>Abstract</h3>
<p>Current audio deepfake detection systems fail to protect specific individuals from targeted voice spoofing attacks.A comprehensive methodology for collecting, curating, and generating high-quality speech datasets specifically designed to protect political figures.</p>
<h3>Salient Features of our Dataset</h3>
<p>✓ Speaker-specific protection </p>
<p>✓ 8-second average duration</p>
<p>✓ Transcription-based segmentation</p>
<p>✓ Natural sentence boundaries </p>
<p>✓ Superior synthesis quality</p>

![image/png](https://cdn-uploads.huggingface.co/production/uploads/661eda10e3e3d83f43de7ba7/uoXLaYiYEcSXZshGSFG05.png)

<h3>Acknowledgement</h3>
<p>
  This work uses only public‑domain or licensed speech sources of public figures. Audio clips are included for research/educational demonstration; rights remain with original owners. Please see the paper for data collection ethics and licensing details.
</p>
<h3>loading_dataset-Using python</h3>
    <p><code>from datasets import load_dataset</code></p>
    <p><code>my_token = "my_token"</code></p>
    <p><code>dataset = load_dataset('issf/famousfigures', cache_dir="path_to_cache",token = my_token)</code></p>
<h3>Cite us : )</h3>

<p><code>@inproceedings{ali25_interspeech,</code></p>
<p><code>  title     = {{Collecting, Curating, and Annotating Good Quality Speech deepfake dataset for Famous Figures: Process and Challenges}},</code></p>
<p><code>  author    = {Hashim Ali and Surya Subramani and Raksha Varahamurthy and Nithin Sai Adupa and Lekha Bollinani and Hafiz Malik},</code></p>
<p><code>  year      = {2025},</code></p>
<p><code>  booktitle = {{Interspeech 2025}},</code></p>
<p><code>  pages     = {3928--3932},</code></p>
<p><code>  doi       = {10.21437/Interspeech.2025-2418},</code></p>
 <p><code> issn      = {2958-1796},</code></p>
<p><code>}</code></p>