Overview

AIVoiceSeparator separates songs into vocal and instrumental stems through a web app or REST API, with optional lyric transcription. It suits creators and developers who need both a browser-based workflow and a straightforward way to integrate separation into an existing stack.

Its broad file support, full-quality free tier and prepaid options make it flexible for occasional use; the single-GPU queue is a meaningful constraint for teams processing batches at scale.

Key features

The Studio pipeline combines BS-Roformer, Mel-Roformer and MDX23C models, with EBU-R128 loudness normalization. The maker reports an SDR of 12.97 dB, roughly 3 dB above the open Demucs baseline. That is a useful quality signal, but it is a maker-reported comparison rather than a guarantee for every source track.

The web app accepts MP3, WAV, M4A, FLAC, OGG, WebM, Opus and MP4/MOV video, up to 100 MB and 20 minutes. Users can submit YouTube or SoundCloud URLs; TikTok clips require saving and uploading the video. Optional Whisper transcription produces SRT, LRC and TXT files and adds about 30 seconds to processing. BPM and key detection are also included on the Free and Pro plans.

The REST API accepts a URL or uploaded file, authenticates with an X-API-Key and returns vocal and instrumental stems. Its marketplace presence on RapidAPI and support for links from YouTube, SoundCloud, TikTok, Bandcamp and Vimeo give developers several ways to feed it media. Teams that need batch throughput beyond the single-GPU queue are directed to contact the maker about dedicated capacity.

The maker says processing runs on its own GPU in Thailand, rather than a third-party cloud. The privacy policy says uploads and results are deleted automatically within 24 hours, audio is not used to train models, and traffic is encrypted via Cloudflare HTTPS. Temporary files sit on a private machine accessible only via Cloudflare Tunnel. The service collects Google email and uid for signed-in accounts, IP addresses for rate limiting and quota counts, and request metadata for debugging, but does not log audio content.

Users are responsible for having the rights to any content they upload or process. That condition matters especially when submitting links to commercially released music.

Pricing

Plan	Price	What it includes
Free	0.00 USD per free	Three songs per month when signed in; full Studio quality; MP3 320, WAV and FLAC; lyrics, BPM and key detection.
Credits	0.50 USD per once	Pay-as-you-go credits, with $1–$20 top-ups for 2–40 songs; credits never expire, no daily limit, and successful separations only are charged.
API credits	0.50 USD per once	Prepaid API credits with no subscription or monthly minimum. Jobs up to 7 minutes cost $0.50, 7–14 minutes cost $1.00, and 14–20 minutes cost $1.50; failed jobs are refunded.
Pro	30.00 USD per month	Billed via Patreon, cancellable anytime; six songs per day, the same Studio quality as Free, MP3 320/WAV/FLAC, lyrics, BPM and key detection.

Free is a capable way to evaluate the full Studio output, but three monthly songs make it a sampler rather than a regular production allowance. Credits suit irregular web use: the minimum top-up is $1, and there is no subscription, expiry or daily cap. API credits use duration-based pricing, which is predictable for developers handling short jobs, while the 20-minute maximum sets a firm ceiling.

Pro makes more sense for an individual who separates songs regularly and can work within six songs a day. Its daily cap may be a poor fit for bursty workloads, and the stated cancellation flexibility does not remove that throughput limit. Teams needing greater batch capacity should ask about dedicated service rather than assume Pro removes the queue constraint.

Platforms

AIVoiceSeparator offers web and API access. The web app handles uploads and supported source links; the REST API fits software that needs to submit files or URLs programmatically. Exports include MP3, WAV and FLAC. The maximum file duration is 20 minutes, and uploads are capped at 100 MB.

Who it's for

This is a practical fit for musicians, creators and small teams that need vocal and instrumental stems without committing to a subscription, and for developers who want an API with per-job pricing. It is less suited to high-volume pipelines that cannot tolerate a single-GPU queue, or workflows that require files longer than 20 minutes. The content-rights responsibility also makes it a poor choice for processing material a user is not authorized to use.

Pros and cons

Pros

Full-quality free access: the three-song monthly allowance includes Studio quality and multiple export formats, so users can assess output without a lower-quality tier.
Flexible pay-as-you-go use: web credits do not expire, failed separations are not charged, and API pricing has no monthly minimum.
Useful input and output coverage: the web workflow accepts multiple audio formats and video, while API input can be a URL or uploaded file.
Clear data-handling commitments: audio is not used for training and files are automatically deleted within 24 hours, according to the privacy policy.

Cons

Limited batch capacity: processing runs on a single-GPU queue, so sustained high-throughput teams may need to arrange dedicated capacity.
Short free allowance: three songs per month is enough to try the service but not much ongoing use.
Hard duration ceiling: the 20-minute maximum excludes longer recordings and mixes.
Rights obligations fall on users: the terms place responsibility for permission to process uploaded content on the user.

Alternatives

AI Voice Isolators is the broader category directory for comparing services. Choose LALAL.AI instead if wider platform coverage matters: it supports mobile, desktop and web alongside API access, with a free relaxed queue limited to 10 minutes and 200 MB per file; free downloads are unavailable.

AudioShake is worth considering when a broader set of platforms or its free starting allowance of 200 credits, MP3 export, 10-minute files and lyric editing better fits the workflow. For metered API processing billed by input duration, BytePlus Voice and Background Audio Separation charges 0.01 USD per once per minute of source duration, with possible separate storage and distribution fees.

Revid Separate Voice API is an API-only alternative at 25.00 credits per once, with a fixed request cost regardless of audio length or file size and limits of 500 MB and 30 minutes. Altered Real-Time Pro is a desktop option for speech enhancement and voice skins, with 20 minutes of daily free processing on macOS and Windows.

Demixr may suit web users who can work within its free 10-minute monthly allowance, 10 MB upload cap and queued processing; its free plan does not provide WAV downloads. Denoise is a mobile app for denoising rather than song stem separation. VocalDrop is a free desktop option for Linux, macOS and Windows with no track limits or account requirement.

Verdict

AIVoiceSeparator is a strong choice for creators who want full-quality stem separation with a usable free tier, or developers who prefer prepaid API credits to a monthly commitment. Its broad input support, optional lyric files and explicit data-handling commitments strengthen the case. Look elsewhere if your workflow depends on more than 20-minute files or sustained batch throughput: the single-GPU queue is the decisive limitation.

	AIVoiceSeparator	clir Focus
Free plan	Yes	Yes
Free trial	Yes	Yes
Paid from	$30/mo	Free
Open source	No	No
Platforms	api, Web	iOS
Free plan	Yes	No
Input types	audio_and_video	audio
Processing mode	upload	realtime
API access	Yes	—
Maximum file duration	20 min	—
Export formats	mp3, wav, flac	—

AIVoiceSeparator vs clir Focus

AIVoiceSeparator

clir Focus