Deploy and Host Whisper ASR Webservice on Railway

Whisper ASR Webservice turns recorded speech into text. Send it an audio or video file over HTTP and it returns a transcript, subtitles in SRT or WebVTT, or just the language it detected. It runs OpenAI's Whisper models on your own instance, so nothing is sent to a third-party transcription API and there is no per-minute bill. This is a community-maintained template; it is not affiliated with the upstream project.

About Hosting Whisper ASR Webservice

Hosting it is simple in shape and has one sharp edge. The shape is a single Python service with no database, no cache and no queue: one container, one volume holding the downloaded model. The sharp edge is that the application has no authentication at all. It defines three routes and none of them checks a credential, because the project is designed to sit on a private network behind something else. On a platform that hands every service a public address, that assumption stops being true the moment the deploy finishes, and an open transcription endpoint is a bill waiting to happen.

This template puts a password in front of the application and binds the application itself to loopback, so the only way in is through the front door. The password is generated for you. One route stays open, the health route, and it returns a single status word so the platform can tell whether the model has finished loading.

Why Deploy Whisper ASR Webservice on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Whisper ASR Webservice on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.

Concretely, this template generates the password, configures the front door, attaches the volume the model is cached in, pins the listening port to the generated domain, caps upload sizes, and points the healthcheck at the one route that stays open, so you get a working and protected speech-to-text API with nothing to fill in.

Common Use Cases

Transcribe interviews, meetings and voice notes without sending the audio to a third party.
Generate subtitles for a video library, in SRT or WebVTT, in one request per file.
Give an application a private speech-to-text endpoint it can call over HTTP.
Detect what language a recording is in before routing it somewhere else.

Dependencies for Whisper ASR Webservice Hosting

A persistent volume for the downloaded model, so a redeploy does not fetch it again.
Enough memory for the chosen model. The default is the base model, which is modest; the large family needs several gigabytes and is slow without a GPU.
Nothing else. No external database, cache or queue.

Deployment Dependencies

Whisper ASR Webservice upstream project: https://github.com/ahmetoner/whisper-asr-webservice
OpenAI Whisper, the model family it runs: https://github.com/openai/whisper
Caddy, used as the authenticating front door: https://caddyserver.com
Template repository, wrapper image and tests: https://github.com/youssefsiam38/whisper-asr-railway
Published image: ghcr.io/youssefsiam38/whisper-asr-railway
The upstream project is MIT licensed and Caddy is Apache-2.0; both are permissive.

Implementation Details

The wrapper adds no application code. It validates the configuration and refuses to start on a missing or short password, an unknown model or engine, WhisperX without a Hugging Face token, a port collision, or an attempt to move the application off loopback. It hashes the password into a proxy configuration that it validates before use, then opens the public port and starts the application behind it. The health route proxies the application's own schema endpoint to decide what to report and discards the response, so it reflects the real state without publishing anything. Both processes are supervised, so if either stops the container stops and the platform restarts it.