Office desktop RTX 3060 as a shared dictation API for every device
whisper.cpp (CUDA) with large-v3 on an office PC, behind nginx + n8n for auth, exposing an OpenAI-compatible endpoint used from desktop, iPhone, iPad and Android.
2-10 peopleTranscriptionUnder $1kSensitivity: High (client/confidential)
Hardware
| Item | Approx. price (as of 2026-10-16) |
|---|---|
| Desktop PC with NVIDIA GeForce RTX 3060 (12GB) | Not stated in source |
| NVIDIA GeForce RTX 3060 12GB (various) (Newegg, related current price) | $449.99 |
Source prices are as stated by the source. Retailer offers: prices checked 2026-10-16, may change. Some links may be affiliate links; we may earn a commission at no extra cost to you. This never changes what we list.
Software stack
- whisper.cpp (CUDA)
- nginx
- n8n
Models used
- Whisper large-v3 (~2.9GB)
Good at
- Short sentences transcribed well under a second; ~2s end to end
- Unlimited use, nothing leaves the network
- Any OpenAI-compatible client works
Weak at
- Four layers (proxy, gateway, n8n, whisper) to maintain
Rough total cost
Not stated in source; can be a normal work PC.
Getting started
- Build whisper.cpp with CUDA and download large-v3.
- Run its server on the private network only.
- Put nginx (TLS) and an auth layer (n8n) in front.
- Expose POST /v1/audio/transcriptions to clients.
Source & credit
Your Own Speech-to-Text Server: Faster Than Typing, Private by Default — Mr Hoelee (blog)
Summarized from the source; quotes are the source author's words. Check the original before buying.