Whisper ASR: Transcribe 22 min audio in 67 sec — CUDA (~20 × real‑time)

Whisper ASR is a CUDA‑accelerated desktop transcription engine built for creators, developers, and anyone who needs real‑time performance locally. Benchmarks: • 22 min audio → 67 sec • Fastest GGML Whisper model • ~20× real‑time on modern NVIDIA GPUs • Fully local, no cloud, no latency This release is part of my broader push toward local‑first AI tools — CUDA‑optimized, offline, and built for speed. Whisper ASR passed Microsoft Store certification and is launching today alongside a limited early‑access drop on Gumroad (25 seats) for users who want to test performance, provide feedback, and help shape the next update. If you try it, I’d love to hear your benchmarks, GPU setups, and any edge‑case audio you throw at it. Your feedback directly influences the next CUDA build. Thanks for checking it out — excited to share this with you.