Customer
Metacast
When Metacast's previous provider stopped completing transcription jobs, its team moved production to Speech is Cheap over a weekend. The API is now Metacast's primary transcription engine.
Customer
Metacast
Use Case
Podcast Transcription
Migration
Over a Weekend
Reported Result
5-6× Faster
The Incident
Metacast builds a podcast app around searchable transcripts and other tools for listening and discovery. Its transcription workflow depended on a serverless provider that started leaving jobs in the starting state. The provider returned no useful error, and Metacast could not restart the affected instances itself.
A customer then reported missing transcripts for recent episodes. By the time the Metacast team discussed the outage publicly, more than 10 premium customers had jobs stuck and its support ticket had gone unanswered for about two days.
The team needed to restore transcription without asking its users to wait for the former provider to recover.
The Migration
Metacast had already spent a few days comparing Speech is Cheap with its existing transcript workflow.
Metacast submitted sample jobs, reviewed transcript quality, and compared processing speed with its existing provider.
The team added a way to choose between its existing provider and Speech is Cheap while it tested the integration.
When the outage arrived on Saturday, Metacast changed production so new transcript jobs used Speech is Cheap.
The Implementation
Metacast kept transcription inside its app and changed the service that processed its prerecorded podcast audio. Speech is Cheap handled the transcript jobs while Metacast continued to own the listening experience and the way transcripts appeared to its users.
Metacast's workflow submits episodes for transcript generation after the source audio exists.
Arnab reports that Speech is Cheap shipped Metacast's requested API changes on the same day.
After the weekend switch, Speech is Cheap became Metacast's primary production path.
The Outcome
These results describe Metacast's workload and its comparison with the provider it replaced.
"When our previous serverless ML provider went down, our premium users' transcription jobs were stuck for days. We hot-fixed Speech is Cheap into production over the weekend, and it totally saved the day. Not only are the transcriptions 5-6x faster with no cold-start delays, but they even shipped custom API changes for us the exact same day. It's now our primary transcription engine."
Arnab's approved testimonial reports a five-to-sixfold speed improvement and no cold-start delays in Metacast's new production path.
Metacast's own year-in-review says the switch let the team do more for users at less cost. The customer did not publish an absolute cost or savings percentage.
Review the limits, workflow, and output options for prerecorded long-form media.
Choose a plan and test it with your prerecorded audio.