v0.5.2

[BrandonWeng]

What's Changed

• Streaming VAD and Speech Segmentation by @BrandonWeng in Streaming VAD and Speech Segmentation #110
• New parakeet-tdt-v3-0.6b models, ~50% faster by @BrandonWeng in New parakeet-tdt-v3-0.6b models, ~50% faster  #113

Conversion Script available here: https://github.com/FluidInference/mobius/tree/main/models/tts/parakeet-tdt-v3-0.6b/coreml

Full Changelog: v0.5.1...v0.5.2

---

This discussion was created from the release v0.5.2.

[BrandonWeng]

A dev reported that the load times are slower on iPhone 13

[BrandonWeng]

Okay so turns out preprocessor needs to probably just run on CPU/GPU, the tensor input is way too big for ANE, a problem that only surfaces on iPhones. It was hidden previously because the fallback happened pretty quickly but with the combined model it became more of a problem