How we compare speech-to-text APIs.

Every provider profile uses the same six core fields, so the first comparison is consistent from one vendor to the next.

These fields describe the integration details that usually decide a shortlist. They are not a single verdict on which provider is best, because the right choice depends on your audio, your languages, and your budget.

Every pricing, language, streaming, diarization, and free-tier value is checked against the provider’s own pricing page and API documentation, not third-party summaries.

Each provider profile lists the primary sources used, so any figure can be traced back to where it came from. When a provider’s live pricing is rendered in a way that resists a clean read, the profile says the number is not fully documented rather than publishing a guess.

The listed price is the cheapest documented pre-recorded rate for a provider’s current flagship model, converted to a per-minute figure so the board can sort. It is not an all-in quote.

Streaming usually costs more than batch, volume tiers cost less, and providers bill in different native units: per minute, per hour, per second, or per token. The same workload can land at a very different real cost, so each profile footnotes the native unit and the exact tier behind the number. For the vocabulary behind these fields, see the speech-to-text glossary.

Accuracy, measured as word error rate, and latency are not published yet, because a fair comparison requires measuring every provider on the same audio and the same benchmark.

Rather than repeat vendor-reported accuracy numbers drawn from different test sets, the directory leaves those fields empty until they can be measured consistently. Treat every figure here as a starting point and confirm current details in the provider’s own docs before you commit.

Start by deciding whether your product needs real-time streaming, speaker separation, broad language coverage, or a particular pricing model.

Then use the provider profiles to build a shortlist and follow the direct website and API documentation links to validate the details that matter for your workload.

Provider pricing, models, and language support change often. Each profile carries the date its fields were last verified and links to the provider’s primary sources so a reader can check for changes.

If a figure looks wrong, the provider’s own pricing or documentation page is the source of truth. Board order follows the rule-based sort described above; no provider buys placement.