Does Buddy record everyone in a meeting?
No. Buddy records the microphone you select. It is designed to remember your own remarks, not remote participants or Windows system audio. Headphones give the cleanest result.
Does my audio leave my computer?
No. Audio capture, Whisper recognition, pronunciation analysis, and speech generation run locally. The default capped Buddy provider sends only the text required for an AI task to DeepSeek. Local Qwen keeps that text on your PC too.
Can I use Buddy without a cloud account?
Yes. Recording, playback, transcription, and local voice need no account. Releases include up to 1,000 replies or 1,000,000 tokens of capped Buddy DeepSeek access, whichever limit is reached first. Download optional local Qwen to keep AI answers and text improvement on your PC.
Which AI models can answer a dialog?
Buddy Trial uses DeepSeek V4 Flash through Buddy's limited gateway and includes up to 1,000 replies or 1,000,000 tokens. Direct DeepSeek uses the same cloud model with your own DeepSeek API key and account billing. On Windows, local Qwen runs the quantized Qwen 3.6 27B model with DFlash acceleration; its roughly 21.5 GB download is slower but keeps dialog text on your computer.
What does the installer download?
The app is self-contained. Larger speech and local AI model packs are installed on demand with progress and completion notifications, so you do not wait for components you may never use.
Which desktop platforms are supported?
Windows 10/11 x64 is stable. The macOS 13+ Apple Silicon build is beta and currently uses the Dock instead of a menu-bar item. Word confidence and lazy per-word phonetics work there; the full local IPA line is currently Windows/Linux-only. Ubuntu 24.04+ x64 is an experimental GTK4 preview; its tray needs a StatusNotifier-compatible desktop. Local Qwen remains Windows-only.
Is pronunciation feedback an accent score?
No. Buddy uses word timing and recognition confidence to flag passages worth reviewing. IPA is a guide, not a judgement of your accent or identity.