集成 UVR 以通过去除背景噪音来提高 Whisper 的准确性 / Integrate UVR for Enhanced Whisper Accuracy by Removing Background Noise

[AlexMa2011]

Is your feature request related to a problem? Please describe.

To improve the accuracy of Whisper in noisy audio environments, it’s crucial to provide it with cleaner audio inputs. The current voice detection models (VAD) are not sufficient for separating vocals effectively in complex audio scenarios.

为了提高在嘈杂音频环境中 Whisper 的准确性，提供更清晰的音频输入至关重要。当前的人声检测模型（VAD）在复杂音频场景中分离人声的效果不够理想。

Describe the solution you’d like

I suggest integrating Ultimate Vocal Remover (UVR), which effectively isolates vocals from mixed audios. This can reduce the interference from background music and ambient noise on Whisper’s voice recognition.

我建议集成 Ultimate Vocal Remover（UVR），该技术能有效从混合音频中隔离人声。这样可以减少背景音乐和环境噪声对 Whisper 语音识别的干扰。

Describe alternatives you’ve considered

While manual preprocessing using separate tools can achieve similar results, it increases the workflow complexity and user effort. An integrated solution within MemoAI would greatly enhance user experience and efficiency.

虽然使用单独工具进行手动预处理可以达到类似的效果，但它增加了工作流程的复杂性和用户的劳动量。在 MemoAI 中集成此解决方案将大大提高用户体验和效率。

Additional context

URLs : UVR
Python package