Skip to content
Database entryUsesLevel 2 of 6 machine operationMachine operation, level 2 of 6Used to calculate 1 entry, level 3Calculated with 1 entry, level 1Database structure

Transcribe a minute of audio

The server electricity used to transcribe a minute of speech into text with a speech recognition model. Projet Celsius estimate.

  • The figure covers speech recognition on the GPU and the server.
  • The electricity range goes from 0.16 Wh to 0.65 Wh, within the scope of this entry.
  • CO2e is calculated from electricity and the worldwide grid; it is not a direct measurement of carbon.
  • Out of scope: the summary, counted in the task, data centre cooling and server manufacturing.
Calculated with
Used to calculate

Origin of the value

The computing centre of the University of Erlangen-Nuremberg measured about 1,000 kWh for 40,000 hours transcribed with Whisper, cooling included. Benchmarks publish measurements per clip of a few seconds; scaled to the minute with an assumed clip length of 8 seconds, they give 0.16 to 0.65 Wh.

Comparison with the other machine operations

See the full list
Transcribe one minutethis entry0.17 gCO2e

Logarithmic axis: each tick is 10 times the previous one.

Characteristics

Unit1 minute of audio
Default locationworld, 458 gCO2e/kWh (Ember 2025, life cycle); 11 other locations to choose from
Countedspeech recognition on the GPU and the server
Not countedthe summary, counted in the task, data centre cooling and server manufacturing
Statusestimated, 3 other published values
All the characteristics identifier, version, rule for the value, sources, licence
IdentifierIA-0018, projetcelsius.com/en/ai/database/IA-0018/
Version0.1 of 2 November 2026
CategoryMachine operations, what the server computes
Choice of the selected valuevalue taken as it stands from its source, without calculation
Sources·
LicenceCC BY 4.0
Projet Celsius, environmental strategy consultancy

We support companies that account for AI in their Bilan Carbone®

Visiter Projet Celsius ↗