Kryton-Protocol/Kryton-Voice-v1

Kryton Voice v1

Kryton Voice v1 is a tensor-compatible linear merge of two Apache-2.0 Kokoro-82M releases: 90% Kokoro v1.0 and 10% Kokoro v0.19. The result is distributed as SafeTensors rather than a pickle-based PyTorch checkpoint.

The included loader, configuration, and af_heart voice passed a CPU forward test that generated the included 1.5-second, 24 kHz smoke-test.wav. Broader voice quality, pronunciation, language, and stability evaluation is still recommended.

Exact source revisions, verified input hashes, output hash, blend weights, and test measurements are in kryton-merge-manifest.json. build_model.py reproduces the merge from the pinned source files.