Doing

The tool provides real‑time voice transcription on macOS, performing all processing locally without sending audio to external services. Users activate listening with a configurable hotkey, speak, and the transcribed text is pasted at the cursor location, with an optional pip that follows the mouse to indicate where the output will appear. It also captures screenshots on demand, attaching them to the transcript to give visual context for downstream AI interactions.

Designed for developers and creators who integrate language models into their workflows, the software works with any AI application that accepts text input. A “YOLO Mode” can automatically submit the transcription after release, streamlining prompt delivery to models such as Claude. The underlying speech engine is based on NVIDIA’s Parakeet‑TDT, and users may supply their own cloud keys for alternative providers if desired.

The product is sold as a one‑time purchase for macOS 14+ on Apple Silicon, with three device activations, free updates, and a refund guarantee. It emphasizes privacy—no account, no subscription, and no audio leaves the machine—while offering fast transcription speeds measured in sub‑second latency.