README.md

metadata

dataset_info:
  - config_name: ALG
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: dialect
        dtype: string
      - name: speaker_id
        dtype: int64
    splits:
      - name: test
        num_bytes: 422983034.212
        num_examples: 1727
    download_size: 427925747
    dataset_size: 422983034.212
  - config_name: EGY
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: dialect
        dtype: string
      - name: speaker_id
        dtype: int64
    splits:
      - name: test
        num_bytes: 248454696.488
        num_examples: 1024
    download_size: 247782272
    dataset_size: 248454696.488
  - config_name: IRQ
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: dialect
        dtype: string
      - name: speaker_id
        dtype: int64
    splits:
      - name: test
        num_bytes: 697811072.846
        num_examples: 2198
    download_size: 713090943
    dataset_size: 697811072.846
  - config_name: MAR
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: dialect
        dtype: string
      - name: speaker_id
        dtype: int64
    splits:
      - name: test
        num_bytes: 201561364.32
        num_examples: 1083
    download_size: 200703785
    dataset_size: 201561364.32
  - config_name: MSA
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: dialect
        dtype: string
      - name: speaker_id
        dtype: int64
    splits:
      - name: test
        num_bytes: 589059863.12
        num_examples: 2902
    download_size: 582828321
    dataset_size: 589059863.12
  - config_name: SAU
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: dialect
        dtype: string
      - name: speaker_id
        dtype: int64
    splits:
      - name: test
        num_bytes: 266749969.375
        num_examples: 1375
    download_size: 267593891
    dataset_size: 266749969.375
  - config_name: UAE
    features:
      - name: audio
        dtype: audio
      - name: text
        dtype: string
      - name: duration
        dtype: float64
      - name: dialect
        dtype: string
      - name: speaker_id
        dtype: int64
    splits:
      - name: test
        num_bytes: 339792881.448
        num_examples: 1048
    download_size: 336006078
    dataset_size: 339792881.448
configs:
  - config_name: ALG
    data_files:
      - split: test
        path: ALG/test-*
  - config_name: EGY
    data_files:
      - split: test
        path: EGY/test-*
  - config_name: IRQ
    data_files:
      - split: test
        path: IRQ/test-*
  - config_name: MAR
    data_files:
      - split: test
        path: MAR/test-*
  - config_name: MSA
    data_files:
      - split: test
        path: MSA/test-*
  - config_name: SAU
    data_files:
      - split: test
        path: SAU/test-*
  - config_name: UAE
    data_files:
      - split: test
        path: UAE/test-*
license: apache-2.0
task_categories:
  - text-to-speech
  - automatic-speech-recognition
language:
  - ar
size_categories:
  - 10K<n<100K

A systematic and standardized benchmark for the multi-dialect Arabic zero-shot TTS task

Paper: "Habibi: Laying the Open-Source Foundation of Unified-Dialectal Arabic Speech Synthesis"
Github: https://github.com/SWivid/Habibi-TTS

Most subsets are licensed under Apache 2.0, except for SAU sourced from the CC-BY-NC-SA-4.0 SADA.