Mark v2 as experimental — known SLSCU val regression

Replace Whisper-LoRA STT with Pathumma audio LLM + new LoRA

Add Open in Colab badge

Add minimum hardware specs section

Add CT2-converted streaming model (int8_float16)

Add STT LoRA adapter (checkpoint-7470 weights + processor files)

Upload README.md with huggingface_hub

initial commit