Skip to content
August 2026

Is Whisper Free? Yes, and Running It Is Not (2026)

The weights cost nothing. The compute, the streaming wrapper and the engineer maintaining it are the price.

Yes. The weights are open and there is no licence fee. That is the answer, and it is the least useful true thing about Whisper's cost.

Free weights are not a free service. You pay for the hardware that runs them, and for a voice agent you pay for something Whisper does not include at all.

Whisper was not built for streaming

A voice agent needs partial results while the caller is still speaking, so it can detect when they have finished a turn. Whisper transcribes a finished audio file. Those are different jobs.

Bridging the gap means a streaming wrapper, turn detection and reconnection handling, all of which somebody has to maintain. By the time that exists, the commercial streaming vendors are usually cheaper than the engineer maintaining it, and that is the arithmetic worth doing rather than the licence one.

What recognition costs when you buy it

OptionPublished rateStreaming?
Whisper, self-runno licence fee, compute is yoursNot natively
Deepgram streaming list$0.0077/minYes, with turn detection
Deepgram, promotional column$0.0065/minYes

Recognition is roughly a sixth of the $0.0465 component floor for a voice agent minute. It is not where the money is. Synthesis, which runs from about $0.014 to $0.085 a minute depending on the voice, moves a bill further than any recognition decision will.

One component worth arguing with. Deepgram's page currently shows $0.0065 with $0.0077 struck through under a limited-time heading with no end date. The archive shows $0.0077 was the pay-as-you-go price and $0.0065 the committed-tier price in a separate column, until the committed price was moved into the pay-as-you-go column and relabelled. The rate never fell. We take the $0.0077 list.

When self-running actually wins

  • Batch transcription of recorded files, where latency does not matter and nobody is waiting on a turn.
  • Volume high enough that owning the compute beats a per-minute rate, which is a real threshold and one you can compute from your own minutes.
  • Data residency or privacy requirements that rule out sending audio to a third party at all.
  • You already run GPUs for something else and the marginal cost of another workload is close to zero.

Outside those four, a managed streaming service is usually the cheaper answer once you price the engineering time honestly. The mistake is comparing a licence fee of zero against a rate of $0.0077 and stopping there.

What we do not publish

We do not compare recognition vendors on accuracy. We have not measured word error rate, and every published benchmark we found was run by a party selling one of the models in it. What we verify is price, and on price the spread is large, checkable and dated.

Questions

Is Whisper really free?
The weights are, with no licence fee. Running it is not: you pay for compute, and for real-time use you also pay for a streaming wrapper that somebody has to maintain. Commercial streaming recognition publishes $0.0065 to $0.0077 a minute, and below moderate volume that is usually cheaper than operating your own.
Can I use Whisper for a real-time voice agent?
Not without additional work. Whisper transcribes finished audio; a voice agent needs partial results as the caller speaks so it can tell when a turn has ended. Adding streaming, turn detection and reconnection is a real engineering project, and the commercial vendors ship it.
Whisper or Deepgram for phone audio?
Deepgram for anything real-time. Its published streaming list is $0.0077 a minute and it handles turn detection, which is the part an agent breaks without. Whisper is the right pick for batch transcription of recorded files at volume, where latency is irrelevant.
How much of a voice agent's cost is recognition?
About a sixth. At published list rates it is roughly $0.0077 of a $0.0465 component floor. Synthesis is the line with real range, running from about $0.014 to $0.085 a minute depending on the voice, and telephony is another 30% at Twilio's published $0.0140.
Which speech-to-text is most accurate?
We do not publish an accuracy ranking, because we have not measured word error rate and every benchmark we found was produced by a vendor with a stake in the result. Price we do verify, with a date on each figure. Treat accuracy claims from any vendor as a claim rather than a measurement.

Tools mentioned

All tools

Sources

Source interests are labelled. Almost everything published about this subject is written by someone selling into it.

More from the blog

All posts