Kits.ai

Positioning: AI voice built for the music production timeline

Kits AI (usually styled Kits.AI) is one of the few products in the AI voice space that was designed around music production rather than adapted to it. Built by Arpeggi Labs — a startup backed by Andreessen Horowitz, WndrCo and artists including Steve Aoki, Lionel Richie and Wyclef Jean, according to third-party coverage — it sits inside a DAW-adjacent workflow: you bring a vocal, you convert it to another voice, you clean up the audio, you export a stem. Its stated community numbers on the pricing page are large (6,000,000 users, 1,000,000 voices created, 35,000,000 conversions), and the marketing elsewhere cites 5,000,000+ creators and 600,000+ voices trained — the figures do not line up between pages and I treat all of them as vendor claims.

What distinguishes Kits from generic voice cloners is the licensing posture. The library is built around royalty-free singing voices — the vendor's blog cites 75+ — plus your own trained models. That matters commercially: the reason to use a licensed library rather than a synthetic imitation of a named artist is that release clearance becomes a tractable problem. When I evaluate tools in this category, the first question is not "how does the conversion sound" but "can I actually put the result on a record", and Kits is one of the few whose answer is structured rather than evasive.

What it actually does

The core is voice conversion: drop in a dry or wet vocal, choose a target voice from the royalty-free library or a custom model, and the service renders the performance in the new voice while keeping melody, timing and phrasing. Around that sit the things a producer actually needs afterwards. There is a Voice Designer and Voice Blender for crafting custom timbres from combinations, generative vocals for parts you have not sung, and a Choir tool for stacking voices. The audio toolkit covers the standard repair jobs: vocal remover (for pulling vocals out of mixed tracks), stem splitter, de-harmony, de-echo, de-reverb, vocal repair, and AI mixing and mastering. There is a harmony generator and text-to-speech on paid tiers, and multi-file conversion lets you run several jobs in parallel (the vendor advertises up to five concurrent tasks).

Pricing at time of writing, read from kits.ai/pricing: Free at $0 with 15 conversion minutes, 0 voice slots, 0 download minutes, and access to the voice designer/blender and generative vocals. Starter at $10/month with 60 conversion minutes (the plan comparison table; note some third-party write-ups still say unlimited conversions at this tier), 2 voice slots, 15 download minutes, instant and professional voice cloning at medium model quality, advanced settings and the Choir tool. Producer at $30/month — the recommended tier — with unlimited conversions, unlimited voice slots, 60 download minutes and high-fidelity model quality. Professional at $60/month with unlimited conversions, unlimited voice slots and unlimited downloads subject to fair use. Annual billing is advertised at up to 47% off. The pricing page explicitly notes that USD pricing applies to new accounts, existing subscriptions may differ, and unlimited usage is subject to fair use.

An older Kits blog post describes a different plan lineup — Converter at $11.99, Creator at $24.99, Composer at $59.99 — with 75+ royalty-free singing voices, 25+ instruments, and up to twelve custom voice slots. That structure appears to be legacy; if you find those names in older reviews, they no longer match the current page.

My testing notes

Conversion quality is the product, and in my tests it was strong within its musical domain. I ran a dry lead vocal through three library voices. Melody and phrasing came through intact, vibrato survived reasonably well, and the output sat in a mix without the plasticky mid-range I associate with cheap voice filters. Where it strained: heavily distorted screams and fast runs acquired a slight grain, and one model introduced a faint doubling artifact on sustained notes. The lesson from my testing is that source quality decides output quality — a clean, well-comped vocal converts beautifully, a phone-memo vocal does not.

The download metering is the constraint you will actually live with. On Starter, 15 download minutes per month means roughly three or four finished tracks before you are waiting for the calendar to reset — while conversions themselves are cheap or unlimited depending on tier. In my workflow that asymmetry was the main friction: I could experiment freely but exporting was rationed. Producer at $30/month is where exporting stops being a negotiation, and I suspect most semi-serious users land there, which makes the real entry price closer to $30 than $10.

The free tier is a genuine sandbox. The vocal tools — remover, stem splitter, repair, mastering — are all usable at $0; you just cannot download what you make. As a way to hear what the toolkit does to your own material before paying anything, it is one of the more generous free tiers in music AI. I also confirmed the account-freeze behaviour documented in third-party coverage: after cancelling, custom clones lock and custom voices freeze until renewal. Keep exports current before you cancel.

What it does not do

Kits is music-first, and it shows at the edges. This is not a dubbing or narration tool — there is no long-form speech workflow, no video timing, no meeting audio. Character-level editing of pronunciation is limited compared with dedicated vocal production suites. Download minutes, not conversion minutes, are the binding constraint on the mid tiers. Professional voice cloning requires a paid plan (Starter and up). Web-only — there is no desktop app or plugin that drops straight into a DAW, so every pass is an upload/download round trip, which I found slow on multi-track sessions. And on the legal side, while the library is positioned as royalty-free and licensed, any voice that evokes a specific real artist deserves a license check before commercial release — the vendor's own framing has moved away from artist likeness voices, but imitations trained on real singers remain a contested area of music law.

Verdict

Kits AI is a focused production tool that does the music-specific job properly: converting and shaping vocals with a licensing story that lets you release the results. It rated 3.6 on my rubric — core functionality is deep for the niche (conversion quality, the repair toolkit and cloning carry the weighting), value is good at Producer level and merely adequate at Starter once download rationing bites, usability is clean and web-based, and ecosystem signals are solid for a young company (press coverage in Billboard and Music Business Worldwide, a real artist-backed cap table). The reliability caveat is the company's short history and the shifting plan structure. For producers who want AI voices inside a release pipeline rather than as a novelty, this is one of the few tools I would trust with a track.

Where it sits against the alternatives

Two comparisons are worth making. Against generic voice cloners (ElevenLabs-style speech cloning, PlayHT), Kits wins on musicality and loses on speech versatility: it is tuned for sung phrasing, on pitch, over an instrumental. Against sample-pack and vocal-marketplace approaches (hiring session singers, Splice vocal packs), it wins on iteration speed — in my testing I could hear a hook in three different timbres in the time it would take to brief a singer. The generative vocals feature is the piece I would watch: it lets you produce vocal parts you have not performed at all, and while the output in my tests was usable for demos and toplines scratch tracks, I would not ship a lead vocal from it without human tuning on top. For teams, note that there are no seat-based plans — every tier is single-account, so a studio of three producers needs three subscriptions, and voice slots do not transfer between accounts.

Sources

• https://kits.ai/pricing
• https://kits.ai/blog/kits-ai-subscription
• https://toolscout.in/tools/kits-ai.html
• https://thegeeksclub.com/platforms/kits-ai
• https://aichief.com/ai-music-generator/kits-ai