Dataset Viewer
Auto-converted to Parquet Duplicate
text
stringlengths
6
247
ملخبن اسمي اخمد وانا اتخدث بلغة العربية
تقنيت ذكائل استنعيية تططور بشكل متسارعا في عصرنا الحالية
Guten Tag, mein Name ist Klaus und ich spreche auf Deutsch.
Die Spracherkennungstechnologie hat sich in den letzten Jahren stark verbessert.
Künstliche Intelligenz verändert die Art und Weise, wie wir kommunizieren.
The quick brown fox jumps over the lazy dog.
Speech recognition technology has advanced significantly in recent years.
Hello, my name is Alex and I am testing the voice data curator.
Artificial intelligence is transforming the way we process language.
Please speak clearly into the microphone for best results.
Hola, me llamo Carlos y estoy hablando en español.
La tecnología de inteligencia artificial avanza rápidamente.
El reconocimiento de voz es una herramienta muy útil hoy en día.
Buenos días, ¿cómo estás tu hoy?
Bonjour, je m'appelle Marie et je parla en français.
La technologie de reconnaissance vocale a fait de grands progrès.
Le ciel est bleu et le soleil brie aujourd'hui.
l'intelligence artificielle transforme notre façon de vivre.
Namaste, miraname Ananya hai, and me Hindi me bhol Rahin.
Bharath ikvivith tao se Bharath deesh hai jaha anik baashai booli jathi hai.
Aug. 25달 17th, 20 natus, 19 natus.
Kritrim Buddha Dima Taka Upiyog Aajhar Kshetrami Hora Ha hai.
YA É C
こんにちは、私の名前はたなかです。日本語でお話します。
人口知能は私たちの生活を大きく変えています
音声に引き切るとはきんねを起きく進ぼしています。
Namaskar, mazenau Priya Aayani Mimara Tithbola Taheen.
Maharashtra is Bharata Thilek Mahatwasi Raja is.
As Zihaman Kux Henah 130 economic struggle
</ Kontakiljinhevo Cumits 3-4宛茶y jeu </ Kontakiljinhevo Cumits 3-4宜室 </ Kontakiljinhevo Cumits 4宛茶y jeu </ Kontakiljinhevo Cumits 5宛茶y jeu 🖤 🖐️ � antennaeet 🖐️ 🖐️ 🖐️ 🖐️ 🗣️ 🖐️ 🖐️ 🖐️ �oline 🤤️ 🖐️ � سوف侍 come Remove Coins 🖐️ 🖑️ 🖐️🖐️ 🖐️ 🖐️ 🖐️ 🖐️ 🖐️ 🖐️

VoiceDataCurator Multilingual Speech Sample

A small multilingual speech dataset curated using the VoiceDataCurator automated quality control pipeline.

Dataset Description

This dataset contains audio clips (MP3) across 8 languages that passed automated quality checks including SNR analysis, silence ratio detection, clipping detection, and duration validation.

Languages

English, Hindi, Arabic, French, Spanish, German, Japanese, Marathi

Quality Criteria Applied

  • Minimum SNR: 10dB
  • Maximum silence ratio: 40%
  • Maximum clipping ratio: 1%
  • Duration: 1–30 seconds
  • Language verified via OpenAI Whisper

Dataset Structure

Each sample includes a .mp3 audio file and a .txt transcript file located in the data/ directory. The dataset_manifest.csv contains per-file quality metrics for all samples (including rejected ones for reference, but only accepted ones are in data/).

How It Was Created

Audio was generated using gTTS and processed through the VoiceDataCurator pipeline. Transcripts were auto-generated using Whisper and should be treated as approximate.

Intended Use

Demonstration dataset for multilingual speech pipeline research and Whisper fine-tuning experiments.

Downloads last month
5