Summary
According to a post on X by @Marktechpost, Zyphra has open-sourced ZONOS3-Mini, a voice-cloning speech model released under the Apache 2.0 license. The mixture-of-experts model has 800 million parameters, with about 100 million active per token. Zyphra reports 80 ms time to first audio on one H100 and says the model scores 66.2 on its own benchmark, ahead of Fish Audio S2 Pro at 61.7. The post's headline says the model runs on phones and laptops, but the snippet does not give device-level results.
AI-assisted summary based on the listed source.
What happened
The Apache 2.0 mixture-of-experts model has 800 million parameters with about 100 million active per token, and Zyphra reports 80 ms time to first audio on one H100. On Zyphra's own benchmark it scores 66.2, ahead of Fish Audio S2 Pro at 61.7. Why it matters: Real-time voice cloning on consumer devices lowers...
Why it matters
Builders and businesses get an open-source option for real-time voice cloning that, according to the post, can run on consumer devices, which could lower costs. The post also notes that easier voice cloning raises the stakes for misuse. Watch for independent benchmark results, since the performance figures so far come from Zyphra's own testing.
What this means for you
Business readers can use this as a signal of where capital, competition, or market attention is moving.
Signal Intelligence
Signal Strength 95%
Technical label WATCH
Public Interest 33
Category MONEY
Reader Depth PRACTICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 8
Novelty Interest Score 94
Consequence Score 34
Curiosity Score 32
Shareability Score 46
Why this is here
VQV surfaced this signal because it is recent, relevant to AI Voice, connected to Scobleizer News Pulse (X).
