Transcription Cost per Approved Hour: REST, Streaming and Human Correction

Two separate billing clocks

xAI advertises STT REST at $0.10 per audio hour and Streaming at $0.20 per hour. Mistral lists Voxtral Mini Transcribe 3 at $0.0035 per minute and a realtime variant at $0.008 per minute. These references apply to named endpoints and versions, not necessarily every feature or supported file. Multiplying the per-minute figures by sixty produces a mathematical hourly equivalent for scenario comparison, but the original row must retain its billed minute unit.

An October 2026 release boundary

Mistral’s release notes date its Voxtral 3 transcription and TTS family to 20 October 2026 public preview. A report describing that model as available earlier in 2026 would be inaccurate. The current price table is a mutable source and does not itself identify the effective timestamp of every SKU. Records therefore keep 20 October as a model release marker while leaving price effective_at unknown unless the document expressly states it.

First-pass speech recognition may need more work

A transcript used for a legal interview, documentary or support investigation may require human correction, diarization, technical vocabulary reconciliation and redaction. The recognition unit rate covers neither guaranteed accuracy nor the labor of reviewing a confusing segment. A reliable cost estimate counts source audio minutes, retranscription attempts, failed upload retries, editing time, confidentiality controls and accessibility exports. Compare endpoints only after the same recording quality and language constraints are established.

A worked direct-recognition estimate

For a hypothetical ten-hour corpus recognized once with xAI REST STT, the public rate gives 10×$0.10 = $1 before corrections. If a tenth of the material must be retranscribed, recognition cost becomes $1.10 assuming that extra hour is billable under the same unit rate. Add editor cost, metadata and platform storage separately. A streaming transcription of ten hours at the stated rate is a different $2 recognition component, not a faster version of a consumer subscription.

Accepted-hour KPI and ethics

Define accepted_hours = audio hours whose transcripts meet a documented quality-control threshold. Compute all recognition charges plus measured correction cost divided by accepted_hours. If no transcript passes, do not report a zero or infinite value as a meaningful comparable tariff; report insufficient accepted output. For regulated audio, obtain consent and document data handling. Version, language and speaker conditions determine workload suitability beyond the pricing scope of this guide.

How to verify the actual invoice

Begin with a small representative workload rather than the cheapest published headline. Record the exact provider and billing channel, API model identifier, currency, account geography, plan and commitment length. The price or quota evidence in this guide is keyed to audio_input_minute. Do not combine proprietary creative credits, token meters or media seconds into a single currency or invent a conversion rate between them. If an enterprise or regional rate is not published, leave its price unknown and obtain a written quote.

Keep separate estimates for failed attempts, optional tools, storage, exports, human quality review and tax. Use source-verified line items from the companion CSV instead of typing permanent numbers into your forecasting formulas. Where an old plan or model was superseded, preserve its original historical record and create an additive later snapshot. This guide does not claim a product test, performance ranking, independent legal assessment or live procurement approval.

Availability, alternatives and editorial limits

Available plans, supported endpoints, trial eligibility, quota reset and contract rights may vary by region and account even when a public website displays a single global currency. Compare an alternative only with the same accepted task, input/output volume, model version and privacy requirements. Vendor-provided marketing claims are not independent evidence of quality. This historical pricing article may be cross-referenced by the model, media or enterprise procurement team, but its purpose is to explain the billing mechanics without duplicating their evaluations or inventing test scores.