Audar-ASR-V1:以阿拉伯语为首的语音识别基础模型,开放权重

1作者: AudarAI大约 1 个月前原帖
大家好, 我们很高兴地发布Audar-ASR-V1,这是AudarAI的首次公开发布。 AudarAI正在构建多语言音频智能,首先专注于阿拉伯语及其多种方言的语音识别。 我们选择从阿拉伯语开始,并不是因为这是我们的最终目标,而是因为我们相信在这里能够产生最大的影响。从这里开始,我们将继续扩展,致力于为全球提供多语言音频智能。 本次发布包括: • Flash 和 Turbo 开放权重模型 • 技术报告 • 评估工具 • GitHub 仓库 • 在线演示 我们非常欢迎研究人员和开发者的反馈。如果您发现问题、有建议,或者想要与您自己的系统进行基准测试,我们非常期待您的来信。 博客 https://www.audarai.com/news/introducing-audar-asr-v1 GitHub https://github.com/AudarAI/Audar-ASR-V1 Hugging Face https://huggingface.co/audarai
查看原文
Hi everyone, We&#x27;re releasing Audar-ASR-V1, the first public release from AudarAI.<p>AudarAI is building multilingual audio intelligence, beginning with speech recognition for Arabic and its many dialects.<p>We start with Arabic—not because it is our destination, but because it is where we believe we can make the greatest impact. From there, we continue expanding toward multilingual audio intelligence for the world.<p>This release includes: • Flash &amp; Turbo open-weight models • Technical report • Evaluation harness • GitHub repository • Online demos<p>We&#x27;d really appreciate feedback from researchers and developers. If you find issues, have suggestions, or want to benchmark against your own systems, we&#x27;d love to hear from you.<p>Blog https:&#x2F;&#x2F;www.audarai.com&#x2F;news&#x2F;introducing-audar-asr-v1<p>GitHub https:&#x2F;&#x2F;github.com&#x2F;AudarAI&#x2F;Audar-ASR-V1<p>Hugging Face https:&#x2F;&#x2F;huggingface.co&#x2F;audarai