My take on Voicebox (jamiepine/voicebox) after running its backend on a Linux box with no GPU:

• ✅ Engine used: Kokoro 82M, with its built-in voice af_heart. It spoke my line clearly in under a second on CPU.
• 🎙️ Kokoro can't copy your voice. It only has preset voices. Cloning needs one of the other engines.
• ❌ My clone try failed: LuxTTS returned just 0.3 seconds of sound from my sample.
• 🧰 Lots of engines to choose from, and one of them (OmniVoice) is marked non-commercial.
• 👍 Good free option for narration with stock voices, run on your own computer.

"The free stock voice works great on a plain CPU. Voice cloning didn't work for me here, so don't cancel ElevenLabs just yet."