One-time purchase · 100% local · AI-agent (MCP) native

Turn any script into natural audio,
read aloud in your own voice.

A Mac voice studio for creators. Clone your voice from a few lines, then narrate 20,000-character scripts that sound like you. Studio-clean audio, entirely on your machine.

macOS (Apple Silicon) · No subscription · Recordings never leave your device
chapter-01.txt · MyVoice
0:00 / 2:22
Speaker similarity 0.87Prosody 84.2~4× realtime

Great narration shouldn't cost you your time or your voice.

Re-recording one line costs an hour

Fix a single awkward take and you re-record, re-level and re-export the whole thing. Again.

Subscriptions pile up, and your voice lives on someone's server

Cloud voice tools bill monthly and keep your voiceprint. You never really own either one.

Noise removal that erases your word endings

Aggressive denoisers swallow the tail of every sentence, and the result sounds processed.

Narrate in your voice, clean audio included.

MyVoice
MyVoicev3 · rollback
Guided line10 / 10
Your voice profile
Read ten guided lines once. Vocast builds a reusable profile, version it, reinforce it, roll back anytime.
script.md · 3 paragraphs
The story begins on a quiet…
By the second act the pace…(editing)
Everything changes when…
18,240 / 20,000 chars
Long-form narration
Paste up to 20,000 characters. Edit any paragraph in place and re-render just that block, not the whole take.
performance transfer
reference take
↓ delivery followed
Performance transfer
Record a reference delivery and the clone follows your pacing, emphasis and pauses, not a flat read.
karaoke view
Read it back in your own voice.
Karaoke lyrics view
Watch each word light up as it plays, and click any word to jump there. Proofing long narration gets fast.
noise removal
StandardResynth
speech loss 0.0% · silence −38 dB
Hybrid noise removal
Studio-clean audio that keeps your word endings. Standard filtering or full resynthesis when you need it.
task center
Narrate chapter-02ETA 40s
Denoise interview.movdone ✓
Task center + ETA
Everything heavy runs in the background with a live ETA, so you keep writing while renders finish.
Live karaoke demo
MyVoice

Paste your script, pick your voice, and hear it read back in a take that actually sounds like you.

We didn't build this on vibes.

PNS
Prosody north-star

Naturalness scored against a human baseline on every render.

0.0%
Word-ending loss gate

Denoise is rejected if it shaves the tail of a sentence.

≥ 0.85
Speaker similarity

Every clone is measured against your real voiceprint.

78+
Automated tests

The whole evaluation stack is guarded in CI, not vibes.

Read the methodology

Your files never leave the machine.

No upload
There is no server. Nothing to send, nothing to leak.
Works offline
After the one-time model download, run it on a plane.
Your files are yours
Voice, scripts and renders stay on your disk. Always.
AI-native · Model Context Protocol

Claude operates this app directly.

Vocast ships a local MCP server, so an AI agent can clone, narrate and denoise for you, entirely on your machine, no network in the loop.

Claude · vocast (local)
Clean this interview and read the intro in my voice.
▸ denoise("interview.mov", mode=standard) → speech loss 0.0% ✓
▸ list_voice_profiles() → picked "MyVoice"
▸ clone_voice(text=…, profile_id="MyVoice") → PNS 84.2
Done, clean cut plus a narrated intro in MyVoice, saved locally.
denoise
clean a file, keep endings
clone_voice
narrate text in a profile
list_voice_profiles
your saved voices
list_history
recent renders
health
engine status
All local stdio, no network, no cloud.

Not a subscription. Buy once, it's yours.

$49one-time
1 year of free updates, optional renewal after
Keep the app and last version forever
Local MCP server included
Use on up to 3 Macs
14-day, no-questions refund
14-day refund · macOS (Apple Silicon)

Questions, answered.

Yes. $49 once. You get a year of free updates and can keep the app plus your last version forever, renewal after that is optional, not required.

On your Mac, in your user folder. There is no account and no server: your voiceprint, scripts and renders never leave the device.

An Apple Silicon Mac (M1 or newer) on macOS 15+. Voice cloning uses on-device Metal acceleration; noise removal works everywhere.

Vocast exposes a local MCP server over stdio. Point Claude, or any MCP-capable agent, at it and it can call denoise, clone_voice and the rest directly.

Only your own voice, or a voice you have explicit consent to use. Cloning is built for creators narrating their own work, not for impersonation.

Every render is scored on speaker similarity, word accuracy, naturalness and a prosody metric, and a render that misses any gate is rejected rather than shipped. The full methodology is on our blog.

14 days, no questions asked. If it doesn't fit your workflow, email support@vocast.me within two weeks for a full refund.

Own your voice.
Keep your words.

Own it for $49, one-time
macOS (Apple Silicon) · one-time purchase · no account