100% local · No cloud · No audio upload

Will it run on
your machine?

Cortility voice input — fully offline, two engines to choose from, three hardware tiers from office laptop to GPU workstation. This page tells you exactly what you need.

2
Engines
6
Models
99
Languages (Whisper)
0
Cloud calls
Self-Check

Which tier am I in?

Three clicks — and you'll know which engine, which model, and which acceleration path matches your machine. Real-time recommendation.

Input
Describe your machine
Values update the result instantly.
Result
Tier: Recommended
CPU
RAM — 8 GB
481632
Enough up to Whisper Medium
GPU
→ What you can use
Heuristic recommendation — the app re-checks your hardware exactly when voice input is activated and unlocks the matching paths.
01 — Engine

Which engine fits you?

Both run locally. Both are freely choosable. Pick by language needs and available hardware.

Default
Whisper
Whisper.net 1.9.0 — ggml
Maximum language coverage. Highest accuracy.
  • 🌍 99 supported languages
  • 📐 Five model sizes — Tiny to Large v3 Turbo
  • ⚡ Optional NVIDIA CUDA acceleration
  • 🛡 Battle-tested open-source engine
CPUVulkanCUDA
Recommended for maximum language coverage & NVIDIA hardware
Optional
Parakeet
sherpa-onnx 1.12.38 — ONNX
Low latency. GPU-agnostic.
  • 🇪🇺 25 European languages, auto-detection
  • ⚡ Low latency, streaming-ready
  • 🎮 Runs on any DX12 GPU without driver hassle
  • 🆕 Current NVIDIA NeMo generation
CPUDirectML
Recommended for AMD/Intel GPUs & lowest latency
02 — Hardware

Three tiers. Clear choice.

From office laptop to GPU workstation. Find the card that best matches your machine.

Tier 1
Minimum requirements
Just enough to run
CPU
x86-64 (pre-AVX fallback Whisper-CPU only)
RAM
4 GB
GPU
— (CPU is enough)
Storage
100 MB + model size
OS
Windows 10 build 19041 (May 2020) or Windows 11
e.g. office laptop, occasional notes
Sweet Spot
Tier 2
Recommended
Comfortable everyday use
CPU
x86-64 with AVX2 (Intel from Haswell, AMD from Excavator)
RAM
8 GB
GPU
DirectX-12-capable GPU (Intel UHD Skylake / AMD GCN 4 / NVIDIA Maxwell+)
Storage
2–3 GB for models
OS
Windows 10/11, current
e.g. daily dictation, mixed models
Tier 3
High-performance
Long dictation sessions, Large models, Smart Modes
CPU
Modern x86-64 with AVX2
RAM
16 GB
GPU
NVIDIA ≥ 4 GB VRAM (CUDA) or modern GPU ≥ 6 GB VRAM (Smart Modes)
Storage
up to 10 GB
OS
Windows 10/11 current + recent GPU driver
e.g. power user with GPU workstation
03 — Models

Six models.
Whisper & Parakeet.

Whisper
OpenAI · 5 models · 99 languages
CPUVulkanCUDA
Tiny
39 M ~75 MB
Very fast notes, lowest accuracy
DEFAULT
Base
74 M ~142 MB
Solid all-rounder for CPU-only devices
Small
244 M ~466 MB
Good accuracy, still CPU-friendly
Medium
769 M ~1,5 GB
High accuracy, GPU recommended
Large v3 Turbo
809 M ~1,6 GB
Best accuracy, GPU strongly recommended
Parakeet
NVIDIA NeMo · 1 model · 25 languages
CPUDirectML
DEFAULT
parakeet-tdt-0.6b-v3
~2,4 GB 4 files

25 European languages with automatic language detection — DE, EN, ES, FR, IT, PT, NL, PL among others. Low latency, streaming-ready.

04 — Smart Modes · Optional

Local LLMs, when dictation should take shape.

Optionally a local language model cleans up, formats, or restructures the transcribed text — “phrase as an email”, “outline as bullet points”.

Note: Smart Modes have no CPU fallback and require a DX12 GPU with sufficient VRAM. Without one they stay disabled — voice input itself keeps working unchanged.
Llama 3.2 3B Instruct
INT4
Size
~3 GB
VRAM
≥ 4 GB
Backend
DirectML
Mistral 7B Instruct
v0.3 INT4
Size
~4,3 GB
VRAM
≥ 6 GB
Backend
DirectML
Privacy

Your voice never leaves this machine.

🔒
Fully local
All models run on your device. No external processing.
☁
No cloud
Audio and text data never leave your machine.
⬇
One-time download
Models are downloaded once from the Menterium CDN — offline-capable afterwards.
05 — FAQ

Answers in one sentence.

Download now.

14-day free trial. Models download once when voice input is activated.

Setup for Windows To the main page