Audio Transcription Cost Calculator
One hour of audio costs $0.36 with Whisper or gpt-4o-transcribe ($0.006/min), $0.18 with gpt-4o-mini-transcribe, and $0.09 of audio input on Gemini 3 Flash Preview (25 tokens/s × 3,600 s = 90,000 tokens × $1/M), per official pricing checked 16 October 2026.
How to use
- Enter the total audio length in minutes. For 2 hours 30 minutes, type 150.
- For Gemini, optionally enter how many transcript tokens you expect per minute of audio, since Gemini bills the transcript as output tokens. Leave 0 to see the audio input cost only.
- Compare the total and the cost per hour for each model.
How it works
OpenAI transcription models are listed with an estimated cost per minute: gpt-transcribe $0.0045, gpt-4o-mini-transcribe $0.003, gpt-4o-transcribe and gpt-4o-transcribe-diarize $0.006, Whisper $0.006, and the live model gpt-live-transcribe $0.017. Cost = minutes × price per minute. Source: developers.openai.com/api/docs/pricing, checked 16 October 2026.
Gemini counts audio as tokens: 32 tokens per second on Gemini 2.5 (ai.google.dev/gemini-api/docs/tokens) and 25 tokens per second on Gemini 3 (docs/media-resolution). One hour is 115,200 tokens on 2.5 and 90,000 on 3.
Gemini audio input prices per million tokens: 3 Flash Preview $1.00, 3.1 Flash-Lite $0.50, 2.5 Flash $1.00, 2.5 Flash-Lite $0.30. 3.1 Pro Preview has no separate audio price, so its $2.00 input price applies up to 200,000 prompt tokens. Cost = audio tokens × audio price + transcript tokens × output price. Source: ai.google.dev/gemini-api/docs/pricing, checked 16 October 2026.
Cost of one hour of audio
| Model | Per hour | Basis |
|---|---|---|
| gpt-4o-mini-transcribe | $0.18 | $0.003 per minute |
| gpt-transcribe | $0.27 | $0.0045 per minute |
| gpt-4o-transcribe / Whisper | $0.36 | $0.006 per minute |
| gpt-live-transcribe | $1.02 | $0.017 per minute |
| Gemini 2.5 Flash-Lite | $0.0346 | 115,200 tokens × $0.30 |
| Gemini 3 Flash Preview | $0.09 | 90,000 tokens × $1.00 |
What the per minute price leaves out
- Gemini also charges for the transcript it writes, at the output price of the model (for example $3.00 per million on 3 Flash Preview). Enter a transcript token estimate to include it.
- Batch and flex discounts, free tiers and taxes are not included.
- OpenAI lists the per minute figure as an estimated cost. Billing for gpt-4o-transcribe models is metered in audio tokens, so a recording with long silences can come out a little different.
FAQ
How much does it cost to transcribe one hour with Whisper?
About $0.36, at $0.006 per minute. gpt-4o-mini-transcribe is half that, $0.18 per hour.
Is Gemini cheaper than Whisper for transcription?
For the audio input, usually yes. One hour on Gemini 3 Flash Preview is 90,000 tokens, or $0.09. Gemini also bills the transcript text as output, which can add a few cents per hour.
Does speaker labeling cost more?
gpt-4o-transcribe-diarize is listed at the same $0.006 per minute as gpt-4o-transcribe.
Why is the live model more expensive?
gpt-live-transcribe streams text while the audio plays and is priced at $0.017 per minute, close to three times the file based models.