Real-time dictation
Global hotkey capture with sub-100ms partials on capable hardware. Text injects into the focused field or copies to clipboard.
DESKTOP APP · v1.0.1
VocalFuse
Local AI transcription and AI note taking powered by OpenAI Whisper Large — dictation, meetings, and batch exports without sending audio to the cloud.
Built on whisper.cpp · Model file ggml-large-v3.bin · From $10/mo
<100msAvg dictation latency
~3 GBWhisper Large on disk
$10–15Monthly plans
100% localAudio stays on-device
Windows 10+
macOS 12+
Ubuntu 22.04+
CUDA optional
Speech to text at desk speed
Hold your hotkey, speak, and watch text land in Slack, Word, VS Code, or any focused window. VocalFuse runs inference on your GPU or CPU — no round-trip to a cloud API, no per-minute billing surprises.
- ✓ Always-on-top pill overlay with 16-bar live visualizer
- ✓ Drag grip, note-mode switch, settings & close controls
- ✓ Batch transcribe files to TXT, DOCX, SRT, VTT
- ✓ Pro adds AI note taker + email summaries
Always-on-top pill · drag grip to reposition
The VocalFuse desktop UI
Pixel-accurate replicas of the Windows client from VocalFuse_cpp — 248×38 pill overlay, six capture states, settings dialog, and system tray menu.
Pill overlay — all states
Idle
Hold to speak
Border #c4a04a · “Hold to speak”
Note ready
Click for notes
Note switch on · “Click for notes”
Recording
Border #e2c06a · 16-bar visualizer
Busy
Working...
Border #9a7a34 · “Working...”
Note taking
Pro batch capture · live meters
Note sending
Sending notes...
SMTP summary dispatch · “Sending notes...”
Settings dialog
460px panel with Product License, Note Taking Mode (email SMTP), and Updates — matching settings.cpp.
Settings
Vocal Fuse
Product License
Required to use Vocal Fuse.
VF-7K2M-9QPL-4XHN
Product key verified.
Note Taking Mode
Record continuously in 3-minute batches until stopped.
Email Delivery
Sent via SMTP when a note session ends.
notes@company.com
smtp.company.com
587
smtp-user
••••••••••••
Updates
Compare your installed version with the latest release.
Installed1.0.1
Latest1.0.1
You are up to date.
Control layout
Right-side controls: 36×16 note switch, settings gear, close button. Hover states shown below.
Hover — settings
Hold to speak
Hover — close
Hold to speak
System tray
Tray context menu: Settings, Minimize, Show, Close — from tray.cpp.
Layout reference: Pill body
#451218, drag grip 18px, logo 18px, Segoe UI labels, note switch 36×16. See the full wireframes in VocalFuse docs.
How VocalFuse works
Subscribe on the web, install the desktop client, activate with your product key, and dictate anywhere.
Subscribe
Pick Basic ($10) or Pro ($15) on the pricing page. Stripe handles billing — cancel anytime.
Download & install
Grab the installer from Downloads. First launch pulls ggml-large-v3.bin if needed.
Activate & configure
Paste your product key, set your hotkey, enable GPU acceleration, and position the pill overlay where you want it.
Dictate & export
Real-time speech to text into any app. Pro users batch-record meetings and export structured notes.
Built for professionals who can't send voice to the cloud
OpenAI Whisper Large
Full Large weights — not a distilled cloud API model. Studio-grade accuracy for accents, jargon, and long-form narration.
Batch transcription
Drop podcasts, interviews, and lectures. Export subtitles and documents without uploading audio anywhere.
AI note taker (Pro)
Capture up to 3 minutes per batch, structure meeting notes locally, and email summaries via SMTP.
Privacy by architecture
Microphone audio never leaves your machine. License checks use HTTPS; voice data does not.
Developer-friendly
License verification API, architecture docs, and local AI guides for teams embedding Fuse intelligence.
Basic vs Pro
| Feature | Basic — $10/mo | Pro — $15/mo |
|---|---|---|
| OpenAI Whisper Large local dictation | ✓ | ✓ |
| Always-on-top pill widget | ✓ | ✓ |
| Hold-to-record hotkey | ✓ | ✓ |
| Batch file transcription (TXT, SRT, VTT, DOCX) | ✓ | ✓ |
| AI note taker mode | — | ✓ |
| 3-minute batch meeting capture | — | ✓ |
| Email summaries via SMTP | — | ✓ |
| Priority updates | — | ✓ |
Who uses VocalFuse?
Developers & writers
Dictate commits, docs, and emails without leaving the IDE. Whisper Large handles technical vocabulary better than lightweight cloud models.
Legal & healthcare
Keep sensitive conversations on-device. No third-party transcription bot joins your calls — audio stays in your trust boundary.
Students & creators
Transcribe lectures and podcasts offline. Pro turns long sessions into structured notes you can export or email.
System requirements
| Operating system | Windows 10/11 (64-bit), macOS 12+, Ubuntu 22.04+ |
|---|---|
| Memory | 16 GB RAM recommended for OpenAI Whisper Large |
| GPU | Optional NVIDIA GPU with CUDA for lowest latency |
| Storage | ~3 GB for ggml-large-v3.bin |
| Model path | C:\VocalFuse\models\ |
| Permissions | Microphone; optional accessibility APIs for text injection |
| Account | Active VocalFuse subscription for downloads and license validation |
Need tuning advice? Read the Local AI Guide for GPU setup, or the full VocalFuse documentation for install and troubleshooting.
Start dictating in minutes
Create your account, subscribe, download VocalFuse.exe, and press your hotkey. No cloud account required beyond Fuse Intelligence.