Nov 14 5 days ago 10:37 Use ONNX-compatible voices.npz (names/values only) voices.npz now contains exactly names and values, as holos-tts-onnx requires. Metadata remains in the CC0 dataset names.json/manifest.json. voices.pt and checksums updated. 10:15 Replace 27 default voices with 29 CC0 Ukrainian voices Voices from Morkon/holos-voices-uk-cc0 (CC0-1.0): 9 male, 14 female, 6 teen. Updates both bundles, legacy tts voices, README and SHA256SUMS. Nov 1 18 days ago 13:46 Update HF model card for new weights, 1000-frame limit, split behavior, re-enrollment 13:46 Upload new voice encoder (fp32 ONNX + PT) and voices.pt 13:45 Upload optional INT8 HolosTTS bundle (official onnx-int8/) 13:45 Upload fp32 HolosTTS bundle (official onnx/, vocab truncated to 181) 10:13 docs: add HolosTTS voice-consent warning 10:13 docs: add HolosTTS voice-consent warning 09:49 Document holos-onnx bundle and HolosTTS attribution 09:49 Add HolosTTS multi-stage ONNX bundle (holos-onnx) Oct 19 20:48 Upload tts/voice_encoder.onnx with huggingface_hub 18:22 Add model card (layout, Gemma license notice) 18:22 Add Gemma-3-270M-uk verbalizer (INT8 MatMul-only QInt8, 226MB) + license notice 13:02 Delete verbalizer/onnx-int8/decoder_with_past_model_quantized.onnx with huggingface_hub 13:02 Delete verbalizer/onnx-int8/decoder_model_quantized.onnx with huggingface_hub 13:01 Upload verbalizer/onnx-int8/decoder_model_merged_quantized.onnx with huggingface_hub Oct 15 15:57 fix accentor: single-file ONNX (1.23MB) replaces broken external-data export (21KB, missing .onnx.data) 09:28 Upload folder using huggingface_hub 09:28 Upload folder using huggingface_hub 09:28 Upload folder using huggingface_hub 09:27 Upload folder using huggingface_hub 08:02 initial commit