your engine / month
per hour
per minute
cheapest / month

OpenAI transcription engines ranked for your volume

Same audio, cheapest first.

Engine$/minuteYour $/mo
⚠️ Estimate. Per-minute prices are reference figures (July 2026) and may change — confirm on OpenAI's pricing page. Audio is billed by duration, not file size. gpt-4o transcription models can also be metered by audio tokens on some tiers; the per-minute figures here are documented list estimates. · Report outdated price →

How Whisper API pricing works

OpenAI's Whisper API charges a flat $0.006 per minute of audio — about $0.36 an hour — and bills by duration, not file size or word count. Usage is rounded to the nearest second, so a 90-second clip costs roughly a cent regardless of how many words it contains; a mostly-silent recording costs the same as a fast, speech-dense one of equal length. To plan spend, take your monthly audio in minutes, multiply by the per-minute rate, and you have the bill. The newer gpt-4o-transcribe sits at a similar ~$0.006/min, while gpt-4o-mini-transcribe comes in around $0.003/min — roughly half — making it the value pick for large batch jobs where you don't need the biggest model.

Whisper vs gpt-4o-transcribe

Because Whisper and gpt-4o-transcribe cost about the same per minute, the decision is mostly about accuracy versus price, not budget. gpt-4o-transcribe generally handles noisy audio, accents and domain vocabulary better; classic Whisper (whisper-1) is a proven, well-understood workhorse. If you're transcribing thousands of clean, clear hours a month and want to shave the bill, gpt-4o-mini-transcribe at ~$0.003/min halves the cost for a modest accuracy trade-off. Test all three on a representative sample of your own audio before committing — word-error rate on your actual content matters far more than a fraction of a cent per minute.

Self-host Whisper vs the API

Open-source Whisper removes the per-minute fee but adds GPU rental, ops and idle time. Below a few hundred hours a month the hosted API almost always wins on both price and simplicity; above that, a GPU running near the clock can beat it. If your transcripts then feed an LLM — summaries, Q&A over calls, search — add that token cost on the LLM token cost calculator, and compare full speech-to-text providers on the transcription cost calculator and speech-to-text cost calculator.

Frequently asked questions

How much does the Whisper API cost?

OpenAI's Whisper API is $0.006 per minute of audio — about $0.36 per hour — billed by duration and rounded to the nearest second. A one-hour recording costs the same whether it's speech-heavy or mostly silence, because you pay for length, not words or file size. Prices are reference estimates; confirm the current rate on OpenAI's pricing page.

Whisper vs gpt-4o-transcribe — which is cheaper?

Whisper and gpt-4o-transcribe both sit around $0.006 per minute, so they cost roughly the same. The newer gpt-4o-mini-transcribe is about half that (~$0.003/min) and is the cheapest for high-volume batch work where you don't need the largest model. gpt-4o-transcribe usually offers stronger accuracy on noisy or accented audio — so it's accuracy versus a lower rate, not a large price gap.

Is Whisper cheaper self-hosted?

Open-source Whisper removes the per-minute fee but you pay for the GPU, ops and idle time instead. Below a few hundred hours a month the hosted API is almost always cheaper and simpler; above that, a well-utilised GPU running near the clock can win. The crossover depends on utilisation — a card transcribing 24/7 amortises well, one used a few hours a day does not.

Related tools & guides

Transcription cost calculator · Speech-to-text cost calculator · LLM token cost calculator · Learn · All AI APIs

Host your project:DigitalOcean — $200 free ↗Hostinger VPS
Transcription CostSpeech-to-Text CostText-to-Speech CostLLM Token Cost