About this site
How we compare providers
SpeechTxt exists to make speech-to-text buying decisions legible. This page explains how the comparisons are put together and what our biases are.
One template for every vendor
Each provider profile uses the same sections in the same order — models, features, deployment, accuracy reporting, pricing, support, pros and cons. If a section is thin for a provider, that gap is itself informative and we leave it visible rather than padding it.
Where the numbers come from
Pricing, feature availability, and language counts come from vendors' own published documentation and pricing pages. Accuracy figures come from published benchmarks and papers, and we name the dataset whenever we quote a number. We do not publish accuracy claims we cannot attribute.
Benchmarks are context, not verdicts
Word Error Rate depends heavily on the audio, the normalisation rules, and the reference transcript. A provider that leads on one dataset can trail on another. We report the conditions alongside every figure so results can be interpreted rather than ranked blindly.
Listing order is editorial
The order vendors appear in listings is set editorially and maintained in one place. Placement is never sold, and no vendor pays to appear, to be removed, or to change how they are described.
Corrections
Provider capabilities and prices change frequently. If something here is out of date or wrong, tell us and we will correct it and update the page's last-reviewed date. Always verify pricing with the provider before purchasing.