interview.network
interview.network · voice clone

Your NotebookLM video,
in your own voice.

Google NotebookLM makes great video — but the whole world gets the same two default voices. Send a 30-second recording; keep the original video, timing and picture, in your voice.

Hear it first

The same NotebookLM video, re-voiced — the default voice, then a cloned voice at quality=good and quality=best (fast | good | best is a knob you pass).

Original · default voice
Cloned · quality=good
Cloned · quality=best
Up and running in five minutes
1
Install decentlan
Windows (native, no WSL) / macOS / Linux — step-by-step guide
$ npm i -g @decentnetwork/lan
2
Open the chat window
Run agentnet ui, then open localhost:8765 — the only terminal command you type.
3
Add the service as a friend
In the chat UI click "Add friend" and paste the address:
72a4A7awtn6BqZnBt3eNhZH9MdYXRkM6Ht5HG2sqoiKR
4
Register your voice (once)
Send a clean 30–60s recording of yourself + /add-voice recording.wav your-name
5
Send a video, get it re-voiced
Send the NotebookLM video to the chat + /voice-clone video.mp4 --ref your-name
6
Collect it in the same chat
Takes ≈ the video length (~7 min video ≈ 7 min). Send /help for everything (incl. YouTube→Chinese dub via /clone-publish).
What it looks like in the chat

Send the video, send one command, get progress and the finished file back — all in the chat window.

agentnet chat — voice clone
Pricing
First job
Free

Videos under 10 minutes — hear your own voice before you decide.

After that
$0.15 / video-minute

Billed off the job ledger, no subscription. Paying adds your key to the allowlist.

Limited
30 min free

First 20 users to register a voice — valid 30 days.

Payment: [USDC / PayPal — TBD, contact the email below]. Your key is your account — nothing to register.

Straight talk
The job runs on a private GPU machine — no big platform touches it; the video never leaves the machine (only the audio track goes into voice conversion); source files are deleted after delivery.
V1 is single-voice: a two-host NotebookLM video gets one voice for both. We'll add multi-voice when enough users need it.
Only your own voice, or a voice you hold written authorization for. Jobs faking someone else's speech are refused and removed from the allowlist.
Use a clean voice sample (30–60s, no background music).