text-to-speech

band: coolmomentum: stable score: 0.006

temperature history

Episodes (6)

Independent testing will determine whether ClearVoice can run OmniVoice multilingual text-to-speech and voice cloning fully offline on supported iPhones and iPads with acceptable quality, latency, and peak memory use.

expiredconvergesscott: medium

Independent use will determine whether Mac Out Loud provides a practical, fully local open-model text-to-speech workflow on macOS without cloud inference or analytics.

expiredknownscott: medium

NineNineSix claims its Apache-2.0 Gepard 1.0 model reaches 68.7 milliseconds median time-to-first-audio and 5.23% WER on one RTX 4090, which would make it a leading low-latency open TTS option on commodity GPU hardware.

expiredconvergesscott: medium

BreezeBlue claims its released Breeze-TTS-2 model delivers frontier-quality text-to-speech in a roughly 7GB locally runnable package, potentially expanding high-quality self-hosted voice generation.

expiredconvergesscott: medium

CrispASR’s maintainers claim their released single-binary C++ runtime can run multilingual speech-recognition and text-to-speech models locally across commodity systems, potentially simplifying self-hosted audio applications.

expiredknownscott: low

VoxGen’s maintainer claims the released Rust and Vulkan runtime makes local VoxCPM2 speech generation practical on AMD hardware without Python, PyTorch, or CUDA dependencies.

expiredknownscott: medium

Trajectory notes

2026-09-04T23:31:00Z: voxgen-vulkan-amd-tts closed (faded) — Scott already maintains a local speech-engine laboratory and self-hosted GPU model zoo, while his Hardware-aware local inference page explicitly treats accelerator placement and runtime policy as engineering concerns. VoxGen could extend t
2026-09-03T00:23:05Z: crispasr-local-audio-runtime closed (faded) — The radar already tracks substantially similar local GGUF audio-runtime developments on “audio.cpp 0.7 local audio arena” and “NeMo-Speech.cpp local stack.” CrispASR could be another candidate for Scott’s local speech-engine laborat
2026-09-01T07:35:06Z: breeze-tts-2-local-release closed (faded) — Breeze-TTS-2 extends Scott’s existing local speech-engine and self-hosted GPU work with a potentially stronger open-weight TTS candidate, creating a concrete benchmark opportunity against Kokoro, Parler-TTS and F5-TTS. The claimed fro
2026-08-28T18:39:17Z: gepard-single-gpu-low-latency-tts closed (faded) — Gepard converges with Scott’s active work on low-latency, locally served TTS and is a concrete Apache-2.0 candidate for benchmarking against Kokoro, Parler-TTS and F5-TTS on his GPU stack. It could change model selection for hi
2026-08-26T21:25:46Z: clearvoice-offline-ios-tts closed (faded) — ClearVoice independently pursues the same offline speech and voice-cloning territory as Scott’s local speech-engine laboratory, while its unresolved iPhone quality, latency, and memory claims map directly to his hardware-aware inferen