docs: README — remove tts-large-fp8/ section (variant retired; runtime NF4 on tts-large/ covers the smaller-card use case)

remove: tts-large-fp8/ — pre-quantized FP8 variant retired (transformers/torchao reload bugs + no CPU/disk spillover; bf16 + runtime NF4 covers the smaller-card use case better)

docs: tts-large-fp8 — correct VRAM requirements (16+ GB), document no-spillover constraint

docs: README — add tts-large-fp8/ section + comparison table

add: tts-large-fp8/preprocessor_config.json (FP8 LM-backbone shard, pre-quantized from aoi-ot/VibeVoice-Large)

add: tts-large-fp8/model.safetensors.index.json (FP8 LM-backbone shard, pre-quantized from aoi-ot/VibeVoice-Large)

add: tts-large-fp8/model-00003-of-00003.safetensors (FP8 LM-backbone shard, pre-quantized from aoi-ot/VibeVoice-Large)

add: tts-large-fp8/model-00002-of-00003.safetensors (FP8 LM-backbone shard, pre-quantized from aoi-ot/VibeVoice-Large)

add: tts-large-fp8/model-00001-of-00003.safetensors (FP8 LM-backbone shard, pre-quantized from aoi-ot/VibeVoice-Large)

add: tts-large-fp8/generation_config.json (FP8 LM-backbone shard, pre-quantized from aoi-ot/VibeVoice-Large)

add: tts-large-fp8/config.json (FP8 LM-backbone shard, pre-quantized from aoi-ot/VibeVoice-Large)

docs: README — add tts-large/ subfolder + aoi-ot attribution

mirror: aoi-ot/VibeVoice-Large → tts-large/

docs: add README — layout, RAI, attribution

mirror: microsoft/VibeVoice-ASR → asr-7b/

mirror: microsoft/VibeVoice-1.5B → tts-1.5b/

mirror: 25 realtime voice presets (KV-cache .pt) from microsoft/VibeVoice GitHub

mirror: microsoft/VibeVoice-Realtime-0.5B → realtime-0.5b/

initial commit