How speech-to-text is priced
Transcription APIs bill per minute of audio. OpenAI Whisper is about $0.006 a minute in 2026, Deepgram and AssemblyAI run $0.004–0.015 depending on model and features, and cloud providers charge $0.016–0.024 for enhanced models. Diarisation (who spoke when), word timestamps and real-time streaming usually cost extra.
Using the calculator
Enter the hours of audio you process in a month and the per-minute price for your provider. The result shows the monthly bill, the cost per hour of audio and how many minutes a $10 budget buys. A podcast network transcribing 50 hours a month at Whisper rates spends about $18, roughly ₹1,500.
Cutting transcription cost
Trim silence before upload, use mono 16 kHz audio, batch files, and consider open-source Whisper on your own hardware above a few hundred hours a month. For Indian languages, check accuracy on a sample first; prices are the same but error rates differ widely between vendors.
Frequently asked questions
How much does Whisper API cost per hour?
About $0.36 an hour at $0.006 a minute.
Is transcription billed on audio length or processing time?
Audio length, rounded to the second or minute depending on the provider.
Does Whisper support Hindi?
Yes, along with most major Indian languages, though accuracy is lower than for English.