Skip to content

Distributed Transcription

cast2md supports distributed transcription, allowing multiple machines to process transcription jobs in parallel with the main server.

Overview

The distributed system uses a pull-based architecture where remote "transcriber nodes" poll the server for work. This is NAT/firewall friendly and provides natural load balancing.

┌──────────────────────┐
│    cast2md Server    │
│  (job coordinator)   │
└──────────┬───────────┘
           │ HTTP API
    ┌──────┼──────┐
    │      │      │
    ▼      ▼      ▼
  Node   Node   RunPod
  (Mac)  (GPU)  (A5000)

Node Types

Type Description Setup
Local machine Mac, Linux, or Windows PC on your network Setup Guide
RunPod GPU pod On-demand cloud GPU for batch processing RunPod Guide

When to Use

Scenario Recommendation
< 10 episodes/day Server only, no nodes needed
10-50 episodes/day 1-2 local nodes (M4 Mac or GPU)
Large backlog (100+) RunPod GPU pods
Ongoing high volume Combination of local nodes + RunPod

Documentation

Page Description
Setup Guide Step-by-step setup for server and nodes
Architecture System design, components, data flow
RunPod GPU Workers On-demand GPU transcription
Performance GPU benchmarks and scaling recommendations