Voice cloning that respects IP

Your voice.
Your hardware.
Your code.

Bloom Voice clones a voice from a 30-second recording. The catch — you buy the source, run it on your machine, and your audio never touches our servers.

$1,500 one-timeNo subscriptionNo cloud

Record your voice ↓See what you get →

Hear it for yourself3 clips

It's my voice. I cloned myself with my own consent.

“Hey, I'm Heathrow, and I built Bloom Voice for one reason. Voice cloning is everywhere now, but almost nobody actually owns their clone.”

Every clip cloned from a single 30-second recording. No editing. No cherry-picked takes. Generated audio destroyed in 24 hours — Bloom Voice ships as code you run on your hardware.

Now try yours↓

Your turn

See exactly what
training feels like.

This is the same 30-second recording flow your customers will go through after you install Bloom Voice. No upload. Nothing leaves your browser. Just a quick proof that the recording format is dead simple.

Hit start, read the suggested script, hit stop.
Audio stays in your browser — there is no upload.
Play it back to hear what real training input sounds like.

Try it on your voiceBrowser-only · nothing uploads

Record 30 seconds.
See exactly what training feels like.

This is the exact format Bloom Voice trains on. Your audio stays in your browser — nothing is uploaded, nothing is stored.

Read this suggested script →▾

Hi, my name is _______. I'm recording this so I can train my own voice clone. I'm sitting in a quiet room, talking the way I'd normally talk to a friend on a phone call. I'm not putting on a voice — this is just me, at a normal pace, reading a few short sentences so the model has enough material to work with. Thanks for listening.

The IP rules →

From the founder

“I cloned myself with this code.
Now you can clone yourself with the same code.”

Hit play to hear what 30 seconds of training audio gets you. That’s Heathrow, the founder of Bloom Voice — generated by the same pipeline that ships in the code drop.

Loom walkthrough — coming soonGet the code →

Heathrow Andrews

Cloned from a 30-second recording

0:000:00

No studio. No retakes. No editing — exactly what your install will produce.

How it works

Three steps. Zero cloud.

You install Bloom Voice once. After that, it’s yours — forever, on your hardware, with no subscription pulling at your wallet.

01
Record 30 seconds
Quiet room, single take, normal voice. Or pick a public-domain voice we ship with — no recording needed.
02
Install on your box
We give you the source, an installer, and a Claude Code or Codex setup script. Apple Silicon, Linux, or any GPU box. 30-day setup support included.
03
Generate forever
Type text. Get audio. As many clips as you want, as often as you want. No per-clip charges. No usage telemetry. No phone-home.

See exactly what lands in your repo→

What you get

A real codebase.
In your repo. In an hour.

When you buy the code drop, you get a tarball that unpacks into a working voice-cloning pipeline. No hosted service. No license-check. No expiring keys. Just files.

~/bloom-voice — bash

$ ./install.sh
✓ Detected Apple Silicon · 16 GB RAM
✓ Installed dependencies (Python 3.12, MLX)
✓ Pulled voice model · ready
$ bloom-voice train ./my-voice.wav
  ▸ cleaning audio …
  ▸ transcribing …
  ▸ training profile …
✓ Profile heathrow ready in 47s
$ bloom-voice say heathrow "Hi, this is me. Cloned."
→ output/clip-001.mp3 (1.8s)
$ that's it. it's yours.
$

What lands in your repotar · 18 MB

bloom-voice/
- └─ README.mdinstall + first clone in 5 minutes
- └─ LICENSEyour-code license, commercial OK
- └─ install.shone-shot setup for macOS / Linux
- └─ src/
  - └─ pipeline/training, demucs, loudnorm
  - └─ api/FastAPI + streaming generate
  - └─ ui/reference Next.js studio UI
  - └─ cli/bloom-voice command
- └─ voices/your trained profiles (gitignored)
- └─ scripts/Claude Code / Codex setup helpers
- └─ docs/API reference, hardware notes, FAQ

Everything you need to clone a voice and ship it inside your product. No phone-home. No license check. No expiring keys.

See what it costs→

Pricing

One price. No tiers.

We don’t want a slice of every clip you generate for the rest of time. Pay once. Run it forever.

Self-hosted code drop

Bloom Voice

The full source, dropped into your project.

$1,500one time

Full source code (Python + TypeScript)
Voice training pipeline
Generation API + reference UI
Installer for Apple Silicon, Linux, GPU
Claude Code / Codex setup script
30 days of setup support
No subscription
No usage fees
No cloud dependency
Yours to ship in your product

Request the code drop →Talk to us first

The math

Bloom Voice pays for itself
in under a year.

Subscription voice APIs are great until you do the math over a product’s real lifespan. Here’s the truth, year by year.

Cost over time

What you pay isn’t the same thing as what you keep.

Bloom Voice is one transaction. Voice-API subscriptions compound for as long as your product is alive.

Solution	1 year	3 years	5 years	Ownership
Bloom Voice One-time. Self-hosted. Your code.	$1,500	$1,500	$1,500	✓ Yours
ElevenLabs Creator Subscription. Cloud-hosted. Their servers. Output stays on their platform.	$1,188	$3,564	$5,940	— Theirs
ElevenLabs Pro Subscription. Higher quota. Still cloud. Per-clip API charges on top.	$3,588	$10,764	$17,940	— Theirs

Subscription pricing as of 2026. Bloom Voice break-even on the Creator tier is roughly 15 months. Against Pro it’s 5 months. After that, every clip you generate is free for as long as your product exists.

FAQ

Common questions.

Why one-time, instead of a subscription?

Because we don't want to be in the middle of every clip you generate. You bought the asset. Run it on your hardware. No bills, no rate limits, no us.

What hardware do I need?

Anything modern with 16GB+ RAM works for inference. Apple Silicon (M1 and up) is recommended for the best out-of-the-box experience. Linux + a consumer GPU is also great.

Does Bloom host any of this?

No. Not your samples, not your generated audio, not even your config. The code drop is a self-contained package — there is no Bloom service to depend on.

Can I clone a celebrity or public figure?

Not without their explicit permission. We won't help you do it, and the code ships with consent prompts baked in. For high-profile use, get it in writing.

Can I use generated audio commercially?

Yes — once you own the code drop, the output is yours. You're responsible for making sure the source voice is licensed appropriately.

What if I want a public-domain voice?

Bloom Voice ships with a public-domain LibriVox profile so you can start generating without recording anything. Great for prototypes and testing.

Request the drop

Ship Bloom Voice
inside your product.

Tell us what you’re building. We’ll send a calendar link to walk through the install, your hardware, and the IP setup before you commit.

Or book a 20-min call →

Your voice.Your hardware.Your code.

See exactly what training feels like.

Record 30 seconds.See exactly what training feels like.

We will not clone a voice you don’t own.

Your voice

Your samples

Your output

“I cloned myself with this code. Now you can clone yourself with the same code.”