Nicheloom visits: 31,534 all-time · 11,218 in the last 7 days

The opportunity tracker for new startups.

Today: 1,041 launches · Niche of the day: giveaway and sweepstakes platforms · See full report →

GitHub

One human voice, 51M params, runs on your CPU

Get picks like this daily. Today's launch numbers, the top launches, AI and tech news, and a daily opportunity pick, straight to your inbox.

Read the full description

Open-weights Chinese TTS: 51.43M params, VITS2, no BERT. Runs on a laptop CPU — no cloud, no API key. One real person's voice, not a stockvoice or a multi-speaker average. MIT. Weights on Hugging Face, code on GitHub.

Visit website ↗ Source listing ↗

PH · launched Oct 14, 2026 · 1 upvote (top 67%)

Similar launches

See all 191 similar launches →

Could you build this?

Partial The inference wrapper and CPU runtime are buildable, but a real version needs a trained VITS2 model that clones one voice well. That requires ML training expertise and clean voice data.

What it would actually take: The stack is a VITS2-style architecture in PyTorch, with ONNX or similar export for CPU inference. The hard part is training: curating a clean single-speaker dataset, tuning the model without a BERT front end, and evaluating quality. This takes speech ML research experience and substantial GPU time for training.

Competitors

Other products that read as similar to this one : 191 launches clear the similarity bar, closest 8 shown.

Attention rank: #101 of 192 (itself plus its competitors, highest first, normalized so YC and Product Hunt are compared fairly).

Launched 342 days after the earliest competitor.

Other launches for this product