TilLabs/kokoro-tts-kazakh

Point download instructions to TilLabs org instead of AnuarSv

Add Anuar Sultanbekov to contributor byline

Add concurrent-worker throughput benchmark

Add CPU benchmark table

Document weight_norm checkpoint format note

Fix weight_norm key format so checkpoint loads correctly with current PyTorch/kokoro (was silently loading with random decoder/predictor weights)

Move contributor credits to top byline, drop descriptions

Embed playable audio samples in model card

Remove Space link (requires PRO subscription, not yet deployed)

Rewrite model card: English, professional, co-authors, CPU-inference focus

Add training hyperparameters table

Add kikiri-kazakh km_m1 checkpoint (selected via held-out WER eval)

initial commit