Product docs
Models & Engines Overview
Map Whisper, realtime models, and Pro advanced ASR to plans, two inference stacks, workflows, and deeper docs in Audio Note.
Models & Engines Overview
Settings
Transcription settings overview screenshot
Terms on this page (hover for details):realtime modelsML Engine
What this page is for
Before picking a model, separate these layers:
- Workflow — file, microphone/app realtime, links, watch folders, and so on.
- In-app model groups — Whisper, realtime models, and Pro ML Engine (advanced ASR models).
- Acceleration — CPU vs GPU (CUDA / Vulkan / CoreML), independent from the model you pick.
This page is the hub that links to detailed docs.
What you pick in the app
| Settings group | Best for | Typical plan | Go deeper |
|---|---|---|---|
| Whisper (+ community) | Files, links, long audio, advanced tuning | Free: six tiers; Standard: 11 official; Pro: + community | Model Usage Recommendations |
| realtime models | Mic/app/global realtime, low RTF; files too | Standard+ (EN/ZH); Pro: all realtime languages | Concepts |
| ML Engine (advanced ASR) | Offline files, multilingual quality | Pro only (Qwen3-ASR, Cohere, Parakeet, …) | Advanced ASR models |
Realtime models and ML Engine are separate groups in Settings: the first favors live capture and low latency; the second (Pro) favors offline file quality.
Plan snapshot (aligned with the desktop app)
| Capability | Free | Standard | Pro |
|---|---|---|---|
| Whisper models | Tiny–Small (six) | 11 official | Official + community |
| Realtime / GPU | No | Yes; GPU; realtime presets | Yes; GPU; custom realtime params |
| Realtime model languages | — | Chinese, English | All realtime languages |
| Advanced ASR (ML Engine tab) | — | — | All available models |
| Record-app target FPS ceiling | — | Higher in-app options | Highest in-app options (actual rate may be lower—see Recording doc) |
| Realtime app video archive | — | — | Yes |
See the pricing page for marketing copy; the in-app plan gate wins if anything differs.
GPU in one paragraph
Whisper file/realtime can use GPU on Standard and Pro (Free is CPU-only). Realtime models usually do not need a discrete GPU. Realtime models and ML Engine models use their own engine and download items, separate from Whisper GPU packages.
Windows NVIDIA users should read NVIDIA GPU & CUDA compatibility and GPU Transcription.
FAQ
Can I download only one model?
Yes—manage downloads in Settings; undownloaded models use no disk space.
Pro: advanced ASR vs Whisper Large?
Whisper has the broadest ecosystem; advanced ASR is an extra Pro offline route. See Advanced ASR models.
Where do parameters live?
Start with Advanced Parameter Transcription, then parameters by engine.
