what you pay
one number, per file: $0.25 per minute of audio. the minute count is rounded up to the nearest minute, billed when the transcript is delivered. if the transcription fails for any reason, you don't pay.
example file sizes
| file | audio length | cost |
|---|---|---|
| 5-minute voice memo | 5 min | $1.25 |
| 30-minute interview | 30 min | $7.50 |
| 50-minute therapy session | 50 min | $12.50 |
| 60-minute deposition | 60 min | $15.00 |
| 90-minute focus group | 90 min | $22.50 |
| 120-minute lecture | 120 min | $30.00 |
what's included at this price
- transcription itself. full audio-to-text on the file you upload, with diarization (speaker labels), word-level timestamps, and the editor to review it in.
- speaker-label bulk fix. relabel "Speaker 1" once and have it propagate to every Speaker 1 row. not gated behind a tier.
- custom vocabulary. per-account list of proper nouns, technical terms, and project-specific words the model should expect. learns across files in the same account.
- every export format. .docx, .srt, .vtt, plain text, JSON. plus the format-specific exports — deposition format, NVivo CSV, Jefferson notation — included, not gated.
- translation. transcribe in one language, export the translation alongside. included.
- summary generation. optional structured summary at the top of the transcript. included.
- private mode. run any file on-device with WebGPU + whisper. nothing uploaded. same price as cloud mode — no premium for the privacy.
- account-level history and search. find a quote across every transcript you've ever produced. search by speaker, by date, by project.
what's not included
- human-grade verbatim transcription. we're an AI tool. for transcripts that need word-perfect accuracy on courtroom-grade audio, hire a court reporter or use rev human ($1.50/min, slower turnaround). we don't replace either.
- real-time / live transcription. we transcribe files. for real-time captioning of meetings, use otter or a meeting-bot product. that's a different job and we don't do it.
- SSO, SCIM, enterprise procurement. individual buyer at launch. firm-level and team-level pricing arrives later.
why no subscription
subscriptions are a vendor's tool, not a buyer's. they smooth revenue for the vendor; they tax buyers who use the product intermittently. for transcription specifically, the buyer pattern is uneven by design — six files in april, zero in may, eleven in june for a research project that ends. paying a monthly minimum during a zero-file month is paying for nothing.
most of the category disagrees and charges a monthly minimum anyway. we don't.
why same price for private mode
on-device transcription costs us less to deliver — there's no server doing the inference, no audio being shipped through our infrastructure. we could charge less for it.
we don't, for two reasons. one: making private mode cheaper than cloud mode would push price-sensitive users toward the private path even when their use case doesn't need it, burning their device's resources for no benefit. two: making cloud mode more expensive would punish users who can't run private mode on their hardware, often the same users with the most sensitive audio. flat pricing keeps the choice neutral.
refunds
if a transcript is unusable — corrupted, in the wrong language, garbled past the point of cleanup — write us within 14 days and we refund the file. no return-form, no escalation path, no retention attempt. one email and a credit back to your card.
we don't refund usable transcripts. the cleanup tax exists in every tool — that's the whole point of the benchmark — and "I had to fix some words" isn't a refund case under any tool's policy, including ours.
how this compares
on the same 30-minute file:
- temi: $7.50, à-la-carte upsells for speaker-label features.
- rev AI: $7.50, polished portal, no on-device.
- rev human: $45, slower turnaround, human-grade accuracy.
- sonix premium: $2.50 per file (at $5/audio hour), plus $22/month subscription, plus per-feature tier upgrades.
- audiohighlight: $6.00 flat. all features. private mode same price.
we're not the cheapest per-file price in the category — sonix gets lower if you have steady volume and pay the subscription. we are the cheapest if you don't pay a subscription, and we don't have a subscription. for the uneven buyer pattern most individuals actually have, we win on price after the second month they don't transcribe.
volume / firm pricing
arrives after launch. for firms, clinics, newsrooms, or research teams expecting steady volume of 50+ hours per month, write hello@audiohighlight.com and tell us your shape. flat pricing stays the default; the customization is in batch upload, account-level access controls, and consolidated billing.