选择音频文件
从设备中选择 MP3、WAV、FLAC 或 M4A 文件。文件最大 50 MB,时长不超过 10 分钟。
AI 音频分离
上传歌曲,移除主唱并生成伴奏,可用于卡拉 OK、排练、翻唱或剪辑。先试听伴奏效果,再下载;分离的人声音轨也会一并提供。
采用 Meta 开发的 Demucs 开源音乐分离模型。
移除人声并制作伴奏
拖入音频文件,或从设备中选择
MP3、WAV、FLAC 或 M4A · 最大 50 MB · 最长 10 分钟
音频会通过安全连接发送至在线 AI 服务进行处理。临时输入和结果文件会在一小时后过期。请仅上传你有权使用的音频。
选择歌曲后,可在这里试听伴奏结果。
选择歌曲后,可在这里试听伴奏结果。
伴奏提取器会将混合歌曲中的主唱与伴奏部分分离,生成适合卡拉 OK、排练或翻唱使用的伴奏音轨。本页以伴奏为主要结果:上传音频后,模型会估算人声与伴奏部分,再由浏览器将鼓、贝斯和其他伴奏合成为一条音轨;你也可以试听或下载分离的人声。生成结果是 AI 对混音的估算,并非还原原始录音工程,因此混响、和声和频率重叠可能留下杂音。如果你主要想提取清唱人声,可使用人声分离页面。

三个简单步骤
从设备中选择 MP3、WAV、FLAC 或 M4A 文件。文件最大 50 MB,时长不超过 10 分钟。
点击生成伴奏并保持页面打开,等待音频分离完成。一次处理会生成无人声混音和独立人声音轨。
在浏览器中试听伴奏音轨,再下载用于卡拉 OK 或练习;如果两种素材都需要,也可以保存人声音轨。
从歌曲中分离主唱,制作适合跟唱、排练或练习乐器的伴奏。
录制翻唱前,先将伴奏作为练习参考。试听处理结果,确认它是否适合你的曲目和用途。
在剪辑中使用无人声音乐底轨,或单独处理歌曲伴奏。本页接收音频文件,暂不支持上传视频。
在编曲草稿或获授权的混音中单独聆听伴奏。分离结果是混音参考,不是原始录音室分轨。
可切换到以人声为主的工具,或将整首歌曲拆分为独立音轨。
了解如何制作伴奏、支持的音频文件、处理方式、隐私和 AI 音频分离的局限。
An AI instrumental remover on this page removes the lead vocal from a song mix to create an instrumental backing track. It also returns the separated vocal, so you can preview or download either result.
Upload an MP3, WAV, FLAC or M4A file and choose Create instrumental track. The tool separates the vocal from the accompaniment, mixes the accompaniment into a backing track, and shows it first for preview and download.
The main result is the instrumental version with the lead vocal reduced. A separate vocal track is also available. If you only want to isolate and work with the singer, use the Vocal Remover page instead.
Yes. This instrumental remover is currently free to use online, with no subscription or account required. To keep the service available, requests may be rate-limited.
Upload MP3, WAV, FLAC or M4A files up to 50 MB and 10 minutes long. Browser and codec support can vary by device.
Yes. It creates an instrumental backing track by separating the lead vocal from the accompaniment. Preview the mix before downloading it for singing or rehearsal.
Yes. The same separation job includes a vocal track as well as the instrumental backing track. The instrumental appears first on this page; use Vocal Remover if an isolated voice is your main goal.
Not at the moment. This page accepts MP3, WAV, FLAC and M4A audio files. Extract the audio from a video first if you have the rights to process it.
Like any AI instrumental remover, it estimates which sounds belong to the voice and accompaniment; it cannot perfectly reconstruct a studio mix. Reverb, backing vocals and overlapping instruments can leave artifacts or vocal bleed.
No. Audio is sent securely to an online service for separation. Temporary input and result files expire after one hour, and audio payloads are not stored in request history.
That depends on the rights to the original recording and composition. Creating a backing track does not grant permission to sample, publish or sell copyrighted music; check your licenses before commercial use.
This page focuses on a combined instrumental backing track and the vocal. For separate drums, bass, guitar, piano and other instrument tracks, use the AI Stem Splitter.