waven.AI
on-device first · open models · cloud fallback

Voice, when and how you want it.

Waven keeps as much on-device as possible, with extra grunt in the cloud for the bigger tasks.

— 1 hr free narration / mo— 1.5 hr free transcription / mo— no credit card
HEAR ITreal synthesis · not a recording
50 / 140
SPEAK ITreal transcription · streamed live
WHAT YOU CAN DO

Dictate, transcribe, narrate, clone.
All in one app, all on open models.

Dictate

System-wide voice input that injects text wherever your cursor lands — email, IDE, Slack, Notion, anywhere. Runs on-device in the Mac app, with cloud streaming for longer or multilingual work. Quietly changes how you use your computer.

Transcribe

Hours of interviews, lectures, and meetings, turned into searchable text with speaker labels.

8 languages·speaker labels

Narrate

Turn scripts into clean voiceovers — audiobooks, podcasts, YouTube, accessibility. Built-in voices in 8 languages, or clone your own in many more.

8 languages·WAV / MP3 / OGG

Clone

Record ten seconds, narrate any script in your own voice. Build a small library of voices to reuse across projects.

10s reference·your voice library
MADE FOR

Creators, developers,
and you.

For creators

Bulk transcription, narration generation, voice cloning — drag-and-drop in the web app, one subscription, free tier covers a podcast a week.

web · bulk · free tier

For developers

Drop voice into anything you're building with one HTTP API. Pay-as-you-go at $0.01 a transcription minute and $0.02 a narration minute, or fold it into a subscription. Same engines, same voices.

http api · pay-as-you-go · webhooks

For you

Dictate notes, transcribe a meeting, narrate a bedtime story in your own voice. The free tier covers what most people need — upgrade only when you outgrow it.

free tier · no credit card · cancel anytime
HOW IT WORKS

On-device first.
Cloud when it has to be.

“On-device” means the Mac app: short English dictation, short transcriptions, and short narrations run right on your Mac. Anything you run in a browser — the web dashboard and the live demos above — uses our cloud. See what routes where.

01 · DESKTOP

Mac app — out now

A real Mac app, not an Electron wrapper: hold a hotkey, speak, and your transcript lands at the cursor. Apple-notarized and on-device first, with Windows and Linux to follow.

02 · ON-DEVICE

Short jobs stay local

We run as much as your laptop can handle on your laptop — keeping your monthly minutes intact and your audio off the internet. Long jobs offload to our GPUs and bill by the minute.

save your minutes
03 · HOME

Home Assistant — out now

An add-on for your home server: keep your wake-word and automations local, and let our GPUs handle cloud-quality dictation and voice — including your cloned voice. None of the setup.

PRICING

Pay monthly or annually.
Cancel anytime.

MonthlyAnnual

Free

Try the whole thing, no card required.

$0/ month
 
  • 1 hr narration / month
  • 1.5 hr transcription / month
  • 1 API key
  • 10 API requests / min
  • Mac app included

Personal

Side projects and solo creators.

$5/ month
 
  • 5 hr narration / month
  • 15 hr transcription / month
  • 2 API keys
  • 20 API requests / min
  • Voice cloning + gallery
  • Mac app included

Studio

Power users and indie teams shipping voice in production.

$49/ month
 
  • 120 hr narration / month
  • 300 hr transcription / month
  • 10 API keys
  • 120 API requests / min
  • Priority queue
  • Voice cloning + gallery
  • Mac app included

Enterprise

For teams with production-scale voice workloads.

Custom
From $2,000 USD / month
  • SLA with response-time commitments
  • Volume discounts
  • Dedicated GPUs
  • Named support
  • Custom contracts

Overage is off by default — plans cap unless you opt in. Past your plan, pay-as-you-go is $0.02 a narration minute and $0.01 a transcription minute.

FAQ

What does it sound like?

Click “Listen to a sample” up top to hear Kokoro or OmniVoice synthesize a real phrase live. Kokoro is the built-in narration engine; OmniVoice powers voice cloning at higher fidelity. We don't ship pre-rendered demo clips — every sample on this page is generated in front of you.

Which models are under the hood?

Narration runs on Kokoro 82M (Apache 2.0) for fast speech in 8 languages — English (US and UK), Spanish, French, Italian, Portuguese, Hindi, Japanese, and Chinese — and OmniVoice for multilingual voice cloning with style direction. Transcription runs English on Parakeet-TDT 0.6B v2 — the same model on-device and in the cloud — with Voxtral-Mini 3B handling multilingual streaming. Speaker labels come from pyannote-3.1, and transcript cleanup and summaries run on Qwen3-4B. All open models — every license and attribution requirement is on the credits page.

Can I use this commercially?

For most output, yes — what you can do with it is set by the license of the model that produced it, not by your plan, and every API response names its model. Narration from Kokoro 82M (Apache 2.0) is cleared for commercial use on every tier, including free. Transcripts from Parakeet-TDT are CC-BY-4.0 — commercial use is fine, but credit NVIDIA when you publish them. The credits page lists every model's license and attribution requirements — check it before you ship. Voice cloning is a paid-tier feature and always requires consent for any voice you clone.

What happens to my audio?

Generated audio files are kept on disk for up to 72 hours so you can re-download them, then deleted automatically; async job results are cached for up to 24 hours. Nothing you upload is used to train our models — not on any tier. If we ever introduce a model-improvement program, it will be opt-in and off by default. Reference audio for voice clones expires automatically after seven days. GDPR export and account deletion are available from the dashboard.

Will the desktop app slow down my computer?

Only short jobs run on-device, and they run in a background process you can throttle or disable. Anything heavier (long-form narration, large transcripts) falls back to our GPUs automatically. The app stays out of the way of whatever else you're doing.

What if I exceed my plan?

Pay-as-you-go overage is off by default — you'll hit your plan cap and get an in-app prompt to upgrade or wait for the monthly reset. Flip overage on with one toggle in billing if you'd rather pay by the minute past your plan ($0.02 USD a narration minute, $0.01 USD a transcription minute) — useful if you want to build on top of our API and incorporate voice into your own projects.

Can I cancel?

Yes, anytime. Cancellation takes effect at the end of your current billing period, and you keep full access until then. No retention pings, no cancellation maze.

Make something today.

No credit card required.