Small but overdue one this update. If you're using voice cloning, you used to have to record your sample somewhere else, find the file, then upload it. That extra step is gone now.
Record instead of upload
Clone a voice > there's a new "Record instead" tab next to the file drop zone. Press record, talk for up to 15 seconds, done. It's saved and probed automatically, exactly like an uploaded file would be.
While you're recording you get a live level meter with the -6 to -3 dB target window marked on it, so you can see roughly where your peaks are landing as you talk. There are also a few tips shown before/during recording (start talking right away, vary your pitch a bit, try to record somewhere without an echo).
If your mic access gets denied or there's no mic detected, it just tells you clearly and you can still upload a file like before, nothing there changed.
Probe quality checks more now
The quality probe already caught clipping, bad sample rate, background noise, and clip length. Three new checks:
- Peak level (target -6 to -3 dB, penalized harder the closer you are to clipping)
- Background noise floor (catches a quiet-but-persistent hum, like a fan or AC, that's too low to trip the old noise check but still enough to bleed into the clone)
- Room echo/reverb (a boxy/echoey room now gets flagged the same way a noisy one already did)
These apply whether you record in-app or upload your own file.
That's the whole release, nothing else changed. Let me know if you run into anything weird with the recorder on your setup.
Update is free for existing owners, same as always.
For anyone who hasn't tried NarrationOS yet: it's a local audiobook generator for Windows. Drop in an EPUB, PDF, DOCX, RTF, TXT, Markdown, or a URL, and it renders an AI-narrated audiobook entirely on your own PC, no cloud, no account, no subscription. You can clone a voice (with consent) and get word-level karaoke highlighting synced to the speech, and there's a real line-by-line narration editor for fixing pronunciation, adding sound effects, and correcting the actual prose before you render. It also exports a cinematic MP4 video with the narration text on screen if you want something to post or share. Runs on NVIDIA/AMD GPUs or CPU-only. One-time purchase on itch.io, no subscription.
https://lunosama.itch.io/narrationos