2026-10-11 16:37 UTC

audio-models

band: coolmomentum: stable score: 0.004
temperature history

Episodes (2)

TontaubeV1’s developers claim their released 2.9B open-weight model enables expressive long-form speech, low-latency local inference, and zero-shot voice cloning in English and German, potentially expanding practical self-hosted TTS workflows.
expiredknownscott: medium
Independent reproduction will determine whether VibeVoice 1.5B can generate useful long-form speech locally on an iPhone at near-real-time speed while using roughly 2.2GB of memory.
expiredknownscott: medium

Trajectory notes