Installing Voice-Pro: Three Days of Walls, and What Finally Worked

I make videos, and I wanted a way to dub them using an open-source tool rather than a paid service, so I set out to install Voice-Pro. It ended up taking me about three days, and I’ll be honest: most of that time was spent stuck rather than making progress. But I did get it working in the end, and the dubbing was worth the effort. What follows is a plain account of each wall I ran into and how I climbed over it, in the hope that it saves someone else a few of those days.

Starting on the wrong platform

I began on a machine with a GPU running Linux, which felt like the natural choice. Then I read more carefully and noticed the project’s own notes said it was tested on Windows, with Linux left unverified. Rather than keep fighting an unsupported setup, I stopped and reinstalled everything on Windows, and it went far more smoothly almost immediately. If you’re a beginner, learn from my detour and start on Windows from the very beginning — it saves a lot of needless struggle.

Installs that froze for minutes at a time

The next obstacle was the installer freezing for long stretches. It turned out that parts of my network were blocking some of the download sources the installer relied on, so it kept hammering away at addresses it could never reach. The fix was to force it to pull everything from the main package source only. For one library that was completely blocked, I downloaded the file by hand and dropped it into place myself. On top of that, I had to manually add a couple of supporting tools — ffmpeg and cuDNN — that the tool expected but didn’t install on its own.

Errors with nothing to read

Even once it was installed, dubbing kept failing, and the worst part was that I couldn’t tell why. There was no visible error to read — just a small yellow notification that flashed for half a second and disappeared. With nothing to go on, I asked Claude Code to surface the hidden error messages and print them out where I could actually see them. That single change made an enormous difference: once the errors were visible, every problem after that became something I could reason about instead of guess at.

The last wall, and a small lesson in how the tool works

After all that, it still refused to dub. I went digging through the now-visible errors and, within about an hour, found that I’d simply left the Reference Audio box empty. The message read 'NoneType' object has no attribute 'read'. The reason makes sense once you understand the underlying model: Voice-Pro uses a voice clone that has no voice of its own and instead copies whatever sample you hand it. With no sample, there was literally nothing for it to read, so it failed. I dropped in a short voice clip, and it finally produced a proper dub.

Once it was running, the tool genuinely delivered. The dubbing worked, and cloning a voice from just a short sample was impressive enough to make the whole ordeal feel worthwhile.

Looking back, the hardest part was never the individual errors — it was the simple fact that I didn’t know how to install this kind of software in the first place, and I got stuck at nearly every step of the way. My takeaway is that Voice-Pro rewards persistence: the setup is steep and unforgiving for beginners, but the result on the other side is real. If you’re willing to push through the install, or to lean on a tool like Claude Code to make the invisible errors visible, it’s a capable dubbing option that costs nothing. Next, I want to spend more time with the voice cloning and see how natural I can get the results to sound.