Whisper hosting
Private speech-to-text behind your own transcription API.
Whisper hosting at AIHostBox means a dedicated, single-tenant Whisper instance on our infrastructure: we install it, pin it to a reviewed release, put it behind TLS on your domain, back it up nightly and keep it patched. Plans start at $12.99 per month and the admin account, the data and any API keys are yours.
Whisper transcribes audio across dozens of languages, and the optimised CPU builds we deploy (faster-whisper) make that practical without a GPU. You get an API endpoint that accepts an audio file and returns text, timestamps and optional translation — the same shape as the commercial transcription APIs, at a flat monthly price.
This ladder buys cores rather than memory, because transcription is CPU-bound. On the Starter plan the base and small models run comfortably faster than real time; the Pro plan handles medium and large-v3-turbo for higher accuracy work.
Whisper hosting plans
Every plan is a dedicated single-tenant instance with NVMe storage, nightly off-box backups, TLS and managed patching. Quarterly, semi-annual and annual terms are discounted at checkout. Model tokens are not included — you bring your own key.
How we run Whisper
The whisper-asr-webservice container (CPU build) on port 9000 running the faster-whisper engine with int8 quantization, with model files cached on a persistent volume.
What we configure for you
- ASR_ENGINE=faster_whisper and ASR_DEVICE=cpu, which defaults ASR_QUANTIZATION to int8
- ASR_MODEL chosen for your plan (base or small on Starter, medium or large-v3-turbo on Pro)
- ASR_MODEL_PATH on a persistent volume so models are not downloaded on every restart
- MODEL_IDLE_TIMEOUT if you would rather free memory between infrequent jobs
- TLS on your subdomain, with the API not left open to the public because the service has no login of its own
Which plan fits
Starter (4 GB, 4 vCPU) runs base and small comfortably, which suits meeting notes, voice memos and clear podcast audio. Pro (8 GB, 6 vCPU) is for medium and large-v3-turbo, where accents, crosstalk or technical vocabulary need the extra accuracy and six cores keep turnaround reasonable.
The honest limits
Everything runs on CPU, so large models approach real-time speed rather than beating it, and one long file will hold the queue. The upstream API is its own REST shape (/asr), not a drop-in copy of OpenAI’s /v1/audio/transcriptions route.
What people run Whisper for
Meeting transcription
Recordings in, searchable text out, nothing sent to a vendor.
Podcast and video subtitles
Timestamped output ready for caption files.
Call analysis
Feed transcripts into a model for summaries and tagging.
Licence and version policy
| Item | Position |
|---|---|
| Upstream licence | MIT |
| Sold as managed hosting | Yes |
| Tenancy | Single tenant — one customer per instance |
| Patching | Critical advisories within 24 hours, high severity within 72 |
Our full position on licences, including the applications we decline to sell, is on the licensing page.
Reading on Whisper
Works well with Whisper
Ollama
Your private LLM server — pull a model, get an API, keep every token on your own box.
from $16.99/moDetails →
Piper TTS
Fast neural text-to-speech that runs comfortably on CPU.
from $5.99/moDetails →
Activepieces
Open automation with hundreds of integrations — and no execution quota.
from $5.99/moDetails →
Node-RED
Wire events, devices and APIs together on a canvas.
from $5.99/moDetails →More Speech, image & video we host
Whisper hosting: your questions
How fast is it without a GPU?
Is my audio stored?
Which Whisper models can I run?
What output formats and options does the API support?
Can it tell speakers apart?
Can I send it video files or phone recordings directly?
Run Whisper without running a server
A dedicated Whisper instance, deployed and maintained by us. Month to month, cancel whenever.