
Getting started was quick. I signed up, verified my email, created an API key in the console, and had a transcript back from the batch English model within a few minutes. New accounts come with free credits, which was enough for me to run a decent amount of my own audio through it before paying for anything.
Accuracy on conversational audio is the main reason I kept using it. I tested call recordings with crosstalk, background noise, people trailing off mid-sentence, and a fair amount of profanity. Modulate handled that better than the other service I tried on the same files.
There's more than one model to choose from. There's a standard batch STT, a faster "vfast" English variant when you care more about turnaround than squeezing out the last bit of accuracy, a multilingual version, and streaming for real-time. They also have synthetic voice detection and AI music detection, which I didn't need but tried out of curiosity and they worked as described. All of it sits behind the same key and the same console.
Pricing is per hour of audio, charged in credits, and each model lists its rate. I could predict what a batch job would cost before running it. The console shows usage by model, remaining balance, and a request log with status per call, so when something failed I could see it without adding my own logging first. You can also request a concurrency increase from the console rather than going through support.
Docs covered what I needed. I didn't have to contact support to get to a working integration. Review collected by and hosted on G2.com.
I need transcripts of real conversational audio and don't want to run my own speech models. Modulate provides that through a straightforward REST API with visible usage and costs. Review collected by and hosted on G2.com.