AI Transcription (Speech-to-Text) API Cost Calculator
Enter how much audio you need transcribed and compare the cost across OpenAI Whisper, Deepgram, AssemblyAI, Google, and Azure, in dollars and rupees, side by side. Each vendor's billing-rounding rule and free tier is built in. No signup, no ads, sources cited below.
How it works
The calculator takes a single audio duration, multiplies it by the number of files in your batch, and prices that volume against every provider's published pay-as-you-go rate. All rates are stored as a single source value, per minute or per hour exactly as the vendor lists it, and converted to a common per-minute basis for the comparison.
- Normalise the duration. Hours are multiplied by 60. Your length is treated as the size of one file; the file count then scales the total:
total minutes = length × 60 (if hours) × files. - Apply each vendor's billing granularity. Whisper rounds to the nearest second; Deepgram, AssemblyAI and Azure bill per second, for all four, billed minutes equal the entered minutes. Google rounds each request up to the next 15-second increment:
billed = ceil(minutes ÷ 0.25) × 0.25per file, which adds a premium when you send many short clips. - Subtract the free tier.When you pick “Per month”, Google's first 60 audio minutes each month are removed before pricing. One-off jobs do not get a monthly free allowance.
- Price it.
cost = billed minutes × rate per minute. Per-hour list prices (AssemblyAI Best and Nano, Azure) are divided by 60 to get the per-minute rate; that derivation is cross-checked in the data module so the per-hour and per-minute views always agree. - Convert and project. LKR is
cost × USD→LKRusing the editable, CBSL-sourced rate. For monthly jobs the annual figure is simply the monthly cost × 12. The effective cost-per-hour column lets you eyeball providers regardless of clip length.
Every rate links to the vendor's own pricing page and carries a last-verified date. The comparison covers pre-recorded, standard tiers only, streaming, real-time, and speaker-diarization surcharges are out of scope so the rows stay apples-to-apples, and enterprise volume discounts are not modelled.
Worked examples
Frequently asked questions
Sources & references
- OpenAI: API pricing (Whisper, whisper-1 per-minute rate)
- Deepgram: pricing (Nova pre-recorded pay-as-you-go)
- AssemblyAI: pricing (Best and Nano per-hour rates)
- Google Cloud: Speech-to-Text pricing (15-second increments, 60 free min/month)
- Microsoft Azure: AI Speech pricing (standard speech-to-text per hour)
- Central Bank of Sri Lanka: daily indicative USD/LKR rate
These rates were last cross-checked against each vendor's pricing page on 2026-06-05. API pricing changes frequently, confirm against the linked source before committing to a provider. The USD→LKR field is editable so you can match the rate your bank gives you.
Related tools
Comments & feedback
Spotted a bug or want an improvement? Tell us, our team reviews every comment, and good ideas get built. Comments are public and anonymous.
Spotted a stale rate, a new provider, or an edge case worth modelling?
Email me at [email protected] , most fixes ship within 24 hours.