Why is dynamic batch cheaper?
It accepts asynchronous processing instead of realtime response. Choose it only when turnaround time fits your workflow.
Official transcription buying routes
Speech-to-text API cost calculator
Convert source audio, channels, reprocessing, latency mode, and budget into a monthly cost and an actionable compatible official API route.
Audio workload
Speech-to-text buying decision
5 official routes
Cheapest compatible official route
Official API · Google Cloud
Speech-to-Text V2 · Dynamic Batch
$3.30
within monthly budget
1,100 billable minutes
Complete workload formula
1,000 min × 1.10 reprocessing × 1 channels × $0.0030/min
This route supports the required batch workflow and has the lowest normalized published rate among the tracked compatible official APIs.
Other compatible routesAccuracy not scored
Scribe v2 via ElevenLabs
Published API usage rate of $0.22 per audio hour. Taxes and add-ons are excluded.
$4.03
Price source GPT Transcribe via OpenAI
Published estimated cost per audio minute.
$4.95
Price source Google documents per-channel billing and rounds each request up to a full second; this aggregate estimate cannot reproduce per-file rounding. Storage, diarization add-ons, accuracy, latency, and taxes are excluded. Rates checked Aug 26, 2026.
Rates checked August 26, 2026
Compare only services that fit the processing mode.
Batch and realtime transcription solve different latency requirements, so the calculator excludes the wrong mode before sorting complete published usage costs. It applies documented per-channel billing only where the provider states it.
Batch and realtime offers are never mixed in one ranking.
Reprocessing is added to the complete monthly audio workload.
Documented Google per-channel billing is included.
The recommendation links separately to its price evidence and provider route.
Questions before purchase
Does the result compare transcription accuracy?
No. Language mix, accents, noise, diarization, and domain vocabulary require a representative accuracy test.
What counts as reprocessing?
Retries caused by bad audio, changed prompts, pipeline failures, or a second pass over the same source minutes.
Need a different AI workload?
Compare video generation or token-based language model costs.