Voicer automates dubbing with Ollama, Qwen3-TTS

The project is less about flashy AI and more about workflow compression: it packages three annoying steps into something a creator or developer can actually rerun.

• –Local models keep the stack private and cheap, which matters more here than raw voice quality.
• –The pronunciation dictionary and sentence batching are the real product features; they make narration usable, not just possible.
• –The biggest caveat is quality control: noisy reference audio gets cloned too, and large batches can drift or hallucinate.
• –Open-source packaging plus a desktop UI lowers the barrier, while the CLI makes it easy to slot into n8n, OpenClaw, or other automation pipelines.