Blog

Cloudflare Workers can't run a multi-minute GPU job or Modal's gRPC SDK — so ScribeToAny treats Modal as a plain HTTP endpoint, fires a fire-and-forget spawn, and lets the GPU box call back over a signed webhook. A walk through the async design, the reconciliation backstop, and the sharp edges.

Captions are for viewers who can't hear the audio and include non-speech sounds; subtitles are for viewers who can't understand the language and translate dialogue only. A clear comparison, with the accessibility rules that decide which you need.

A plain explanation of Word Error Rate — how transcription accuracy is actually measured, what the numbers mean, why even humans aren't at 100%, and the factors that move accuracy up or down.

A reference guide to the file formats transcription produces — SRT, VTT, TXT, TSV, CSV, JSON, plus ASS, SSA, SMI and TTML — with real examples of each and a table of when to use which.

Speaker diarization is the step that splits a recording by speaker — turning a wall of text into a labelled dialogue. What it is, how it differs from speaker recognition, how it works, and where it struggles.

Turn a podcast episode into clean, timestamped show notes and captions — transcribe the audio, then export SRT, VTT or plain text. Free plan, no credit card.