Speech-to-text APIs

Speech-to-text APIs turn audio into text: recorded calls, meetings, podcasts and videos sent as batch jobs, or live audio streamed while people speak. Developers use them to add transcripts, captions, call analytics and voice features to their products. They differ in how they price an hour of audio, which features cost extra (speaker labels, custom vocabulary, redaction), whether your audio is used to train their models, how long they keep it, and where it can be processed.

We compare them on the same six indicators, read from each vendor's pricing, docs, data-usage, trust and legal pages:

• Pricing clarity: whether batch and streaming rates are published and can be bought online, whether some rates are promotional, and whether the pages leave prices out or contradict each other.
• Free credit: the sign-up credit or free tier, converted into hours of the reference workload, and whether it needs a card or expires.
• Price for 500 hours: the monthly cost of 500 hours of pre-recorded English with speaker labels and word timestamps as batch jobs, at pay-as-you-go list prices, with the arithmetic shown.
• Transcription features: streaming, streaming speaker labels, custom vocabulary, PII redaction, language detection, multichannel audio and the number of languages, and whether each is included or costs extra.
• Data use and retention: whether customer audio trains the vendor's models by default, how long it is kept, whether it can be deleted, and whether a DPA comes with the terms.
• Compliance and deployment: SOC 2 Type II, ISO 27001, a HIPAA agreement, EU processing, self-hosting and console single sign-on.

Accuracy, word error rate and speed depend on your audio and are not compared here; we do not publish them. Each review says which pages it was written from and when. Where we have run a product hands-on, the evidence is on its page.