What does it cost to transcribe an hour of audio?
Between $0.025 and $0.46 per hour for batch transcription, depending on the vendor. Add speaker diarization and the spread widens, because most providers bill it as a separate add-on. Orchard is $0.025 per hour, or $0.036 per hour with diarization included — the cheapest of the providers below, with or without it.
Every figure here was read off the vendor's own pricing page, not copied from another comparison article. Where a vendor publishes per minute, the hourly number is derived and the original is shown next to it so you can check the arithmetic.
Verified 2026-08-10 · re-checked monthly
Cost at your volume
| Provider | $/hr | Monthly at 1,000 h | vs Orchard |
|---|---|---|---|
| Orchardorchard-stt-v1 | 0.036 | $36 | baseline |
| Sonioxasync | 0.100 | $100 | 2.8×+$64 |
| AssemblyAIUniversal-2 | 0.170 | $170 | 4.7×+$134 |
| AssemblyAIUniversal-3.5 Pro | 0.230 | $230 | 6.4×+$194 |
| DeepgramNova-3 (pre-recorded) | 0.582 | $582 | 16.2×+$546 |
Not shown with diarization on: Groq whisper-large-v3-turbo, Groq whisper-large-v3, OpenAI gpt-4o-mini-transcribe, OpenAI Whisper / gpt-4o-transcribe. These providers do not offer speaker diarization, so there is no like-for-like number to compare — listing their transcription-only price next to a diarized one would flatter us for the wrong reason.
Published rates, per hour of audio
| Provider | Transcription | Diarization | Total | As published |
|---|---|---|---|---|
| Orchardorchard-stt-v1 | $0.025 | +$0.011 | $0.036 | $0.00042 / min |
| Groqwhisper-large-v3-turbo | $0.040 | not offered | — | $0.04 / hr |
| Sonioxasync | $0.100 | included | $0.100 | $0.10 / hr |
| Groqwhisper-large-v3 | $0.111 | not offered | — | $0.111 / hr |
| AssemblyAIUniversal-2 | $0.150 | +$0.020 | $0.170 | $0.15 / hr |
| OpenAIgpt-4o-mini-transcribe | $0.180 | not offered | — | $0.003 / min |
| AssemblyAIUniversal-3.5 Pro | $0.210 | +$0.020 | $0.230 | $0.21 / hr |
| OpenAIWhisper / gpt-4o-transcribe | $0.360 | not offered | — | $0.006 / min |
| DeepgramNova-3 (pre-recorded) | $0.462 | +$0.120 | $0.582 | $0.0077 / min |
Orchard Diarization brings the total to $0.036/hr. Word-level speaker attribution, not segment-level.
Groq No diarization offered. Cheapest transcription-only option after Orchard.
Soniox Diarization, language ID and formatting bundled into the hourly rate.
OpenAI No diarization in the transcription API.
Deepgram Pay-as-you-go rate. Diarization is a $0.0020/min add-on. Committed-volume tiers are lower.
Diarization is where advertised prices stop matching real bills
Almost every workload that justifies transcription has two people in it: a call centre conversation, a sales call, a doctor and a patient, a hearing. For those, a transcript without speaker labels is not a cheaper version of the product — it is a different, unusable one.
That is why the base rate is the wrong number to shop on. Deepgram adds $0.120 per hour for diarization, which is more than three times what Orchard charges for transcription and diarization together. AssemblyAI is gentler at $0.020, and Soniox bundles it like we do.
The second thing the sticker price hides is quality of attribution. Assigning whole segments to whichever speaker overlaps most is cheap and wrong the moment two people talk over each other, which on real calls is constantly. Orchard resolves each word against its own timestamp — a change that took our attribution accuracy from 62.5% to 96.9% on real client audio, while word error rate did not move at all. WER was never the failing metric.
Check it against your own audio
Orchard is drop-in compatible with the OpenAI Whisper API, so testing costs one line: point your existing SDK at https://api.orchardrun.com/v1. The free tier covers 500 minutes a month with no card.
Spotted a price that has moved? Tell us and we will correct it. This table is only useful if it is right, and vendors change pricing more often than comparison articles get updated.