Request evaluation of a speech recognition model
Cloning Voice tokoh Indonesia - Bahasa Indonesia
Generate customized audio from text using a voice sample
Generate high-quality speech from text with specified emotion and voice
Transcribe YouTube videos to text
✨[With v1.0.0] Accelerated TTS on Kokoro-82M
Generate speech from text with adjustable speed
Simple Space for the Kokoro Model
Generate audio from text input
WebGPU text-to-Speech powered by OuteTTS and Transformers.js
Belarusian TTS
Transcribe audio with emotions and events
Open ASR Leaderboard is a platform designed to evaluate and benchmark speech recognition models. It provides a centralized location for developers and researchers to assess the performance of their automatic speech recognition (ASR) systems against established standards and compare them with other models.
• Comprehensive evaluation metrics: The leaderboard provides detailed performance metrics, including word error rate (WER), character error rate (CER), and real-time factor (RTF).
• Multi-language support: It supports evaluation across multiple languages and accents, making it a versatile tool for diverse datasets.
• Benchmark datasets: Access to standardized test datasets for consistent and fair model comparison.
• Customizable evaluation: Users can define specific test scenarios or use predefined configurations.
• Visualization tools: Results are presented in interactive charts and tables for easy analysis.
• Community collaboration: A forum for sharing insights, best practices, and model improvements.
What types of speech recognition models can I evaluate?
You can evaluate any automatic speech recognition model, including deep learning-based models, traditional HMM-based systems, or hybrid approaches.
How often are the leaderboards updated?
The leaderboards are updated regularly as new models are submitted and evaluated. Updates are typically announced in the community forum.
Can I use custom datasets for evaluation?
Yes, you can upload custom test datasets for evaluation, provided they meet the platform's formatting requirements.