Datasets:

Firoj112
/

tts_test

language:
  - en
  - ne
tags:
  - tts
  - audio
  - speech-synthesis
  - wiseai
license: mit
dtaset_info:
  features:
    - name: text
      dtype: string
    - name: audio
      dtype: audio
    - name: language
      dtype: string
    - name: category
      dtype: string
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/*

WiseAI TTS Evaluation Dataset

This dataset contains generated audio samples from the WiseAI TTS system. It is designed for evaluating the performance and quality of the Text-to-Speech engine across English, Nepali, and Code-mixed (Nepali-English) inputs.

Dataset Details

This is the reference dataset generated after running all tests and reports presented in the TTS Evaluation report.

Source Data

The text inputs were sourced from the following test file in the WiseAI codebase: test/test_dataset/jsonl_format/asr_test_data.jsonl

Data Fields

audio: The generated audio file (.wav format, 24kHz).
text: The input text used to generate the audio.
language: The language tag of the input text (english, nepali, or codemix).
- Note: codemix inputs were processed using the Nepali model.
category: The category of the intent/domain (e.g., banking, general).

Usage

from datasets import load_dataset
dataset = load_dataset("Firoj112/tts_test")
# Listen to a sample
print(dataset['train'][0]['text'])