Universal-2
$0.15/hr
Pre-recorded STT
Widest language list
Universal-3.5 Pro
$0.21/hr
Native code switching
Shorter language list
Streaming
$0.15/hr
Live audio, base rate
Pro Realtime is $0.45/hr
Voice Agent
$4.50/hr
STT, LLM, and TTS
billed per second
Also Sync API at $0.45 an hour and Dictation at $0.62. AssemblyAI has no contracts, minimums, or monthly fee, and add-ons bill per hour on top of the base rate.
Key takeaways
- AssemblyAI bills per second with no plan fee: $0.15 an hour on Universal-2 and $0.21 on Universal-3.5 Pro for pre-recorded audio.
- Streaming starts at $0.15 an hour, Pro realtime is $0.45, and the Voice Agent API is a flat $4.50 an hour.
- Add-ons such as diarization, keyterms prompting, and medical mode bill on top, and multichannel audio is charged per channel.
- New accounts get $50 in free credit with no credit card, which AssemblyAI states covers about 185 hours of pre-recorded transcription.
- PyAI Hear is $0.001/min, so 100 hours of audio costs $6 on PyAI against $15 on Universal-2.
AssemblyAI pricing by product: models, streaming, and Voice Agent
AssemblyAI does not sell monthly plans. You pay per second of audio at a rate set by the model, and you only pay for what you process. The table below lists AssemblyAI's main products and what each one costs.
The high accuracy async model, with the widest language list in the AssemblyAI lineup. The default for most batch transcription.
The flagship async model, with native code switching and speaker diarization, on a shorter language list than Universal-2.
Universal-Streaming is the cost efficient live model. Universal-3.5 Pro Realtime is the highest accuracy live model at $0.45 an hour. Concurrency scales automatically at no extra cost.
The Sync API returns a full transcript in one low latency call for short clips. The Dictation API is tuned for short-form dictation to text.
Speech to text, turn and interruption detection, an agent language model, text to speech, and hosting over a single WebSocket, billed per second of connected time.
Volume discounts, custom rate limits, higher concurrency, and AWS Marketplace billing for large workloads, priced by agreement.
Billing is per second, so you are not rounded up to the minute. Multichannel audio is billed per channel, which means a one hour stereo file counts as two hours.
How much AssemblyAI really costs per month
Because AssemblyAI is usage based, your monthly bill is hours of audio times the model rate, plus any add-ons. AssemblyAI has no subscription, so the variables are volume, model, and which features you switch on.
Which AssemblyAI model and pay as you go plan fits your use case
The right AssemblyAI model comes down to whether you transcribe recorded files or live audio, how much accuracy and language coverage you need, and whether you want a full voice agent rather than raw transcription. Since every model bills per hour with no plan, you are matching the model to the job, not committing to a tier.
At $0.15 an hour with the widest AssemblyAI language list, Universal-2 is the value pick for podcasts, meetings, and call recordings where language coverage matters more than the last point of accuracy.
At $0.21 an hour Universal-3.5 Pro adds native code switching and diarization, for medical, legal, or multi speaker audio where errors are costly, within its shorter language list.
Universal-Streaming at $0.15 an hour suits live captions and note takers. Pro Realtime at $0.45 is for live work that needs the highest accuracy, with auto scaling concurrency.
At a flat $4.50 an hour, the AssemblyAI Voice Agent API bundles transcription, an agent model, text to speech, and hosting, so you ship a voice agent without wiring the layers together yourself.
Volume discounts, custom concurrency, and AWS Marketplace billing for teams processing large audio libraries on a committed agreement.
If you only need raw transcription, the pre-recorded models are far cheaper per hour than the Voice Agent API, which bundles a language model and text to speech you may not use. Match the product to the job.
AssemblyAI pricing calculator: estimate your monthly cost
AssemblyAI figure = hours times the model rate per hour. PyAI figure = PyAI Hear at $0.001/min, about $0.06 an hour, billed per minute with no plan. Add-ons and multichannel are left out.
AssemblyAI hidden costs: add-ons, multichannel, and streaming
- Add-ons stack on top: diarization at about $0.02 an hour, keyterms prompting at about $0.05, and medical mode at about $0.15 bill on top of the base rate, so a feature rich job costs more than the headline model rate.
- Multichannel doubles the bill: stereo and multi track audio is billed per channel, so a one hour two channel recording is charged as two hours of transcription.
- Streaming and Sync cost more: the highest accuracy live model is $0.45 an hour and the Sync API is $0.45, three times the $0.15 pre-recorded rate, so realtime work is priced above batch.
- Voice Agent bundles layers you may not need: the $4.50 an hour Voice Agent API includes a language model and text to speech, which is good value for an agent but overkill if you only want transcription.
- Free credit runs out: the $50 credit covers about 185 hours pre-recorded by AssemblyAI's own estimate, then you move to pay as you go, so budget for the point where steady volume passes the free tier.
- LLM Gateway is billed by token: if you add summarization or analysis through the LLM Gateway, you pay per input and output token on top of transcription, separate from the per hour audio rate.
AssemblyAI does not hide these charges, but the charges are easy to miss when you compare only the base per hour rate. The usage model rewards matching the model and add-ons to the job.
AssemblyAI reviews: pricing and product
AssemblyAI is rated 4.6 on G2 from 125 reviews as of October 2026. The score and themes below come from those reviews. The quotes under them are real and attributed.
4.6 on G2 (125 reviews)
Scores and themes are from G2 (125 reviews), checked October 2026.
What users say about AssemblyAI pricing
Pricing is fair, and they give you a generous free trial to test your integration without having to pay.
Pricing can become expensive at scale, especially for apps with high audio usage.
What users love
It was quick for our team to integrate, and it delivers accurate transcription results even with long audio files and conversations involving multiple speakers.
Where users struggle
Transcription accuracy occasionally drops with poor audio quality or strong background noise.
On G2, AssemblyAI is most praised for accuracy and ease of use, and the most common criticisms are limited language support and pricing issues.
AssemblyAI vs PyAI: cost per hour and features compared
AssemblyAI and PyAI both bill only for what you use, so the difference is the rate and the shape of the product. AssemblyAI prices per hour of audio and leads on model accuracy, audio intelligence, and language coverage. PyAI prices per minute and focuses on low cost transcription with phone ready audio and text to speech on the same bill. The table lines them up for raw speech to text.
| AssemblyAI | PyAI | |
|---|---|---|
| Pricing model | Pay as you go, rates quoted per hour | ✓ PyAI bills per minute, no plan fee |
| Pre-recorded transcription | $0.15 to $0.21 an hour | ✓ PyAI Hear $0.001/min, about $0.06 an hour |
| Monthly fee | $0, usage based | $0, PyAI bills only the minutes used |
| Free to start | $50 credit, about 185 hours per AssemblyAI | ✓ $50 credit, no credit card |
| Model accuracy and audio intelligence | ✓ Universal-3.5 Pro, diarization, LeMUR analysis | Transcription focused |
| Language coverage | ✓ Widest language list on Universal-2, shorter list on Universal-3.5 Pro | Multiple languages for transcription |
| Output for phone systems | Transcripts and JSON | ✓ WAV, PCM, or G.711, phone ready |
| Text to speech on the same bill | In the Voice Agent API only | ✓ Speak $0.04/min and Cast $0.02/min |
| Voice agent, all in | ✓ Voice Agent API $4.50 an hour | Agents $0.08/min, about $4.80 an hour |
| G2 rating | ✓ 4.6/5 (125 reviews) | No G2 profile yet |
Choose AssemblyAI if you want
- Universal-3.5 Pro with native code switching and speaker diarization
- Summarization and LeMUR analysis on top of transcripts
- The widest AssemblyAI language list on Universal-2
- A flat $4.50 an hour Voice Agent API from the same vendor
Choose PyAI if you want
- Transcription at $0.001/min, about $0.06 an hour
- Transcripts and synthesized audio that plug into phone systems, with output in WAV, PCM, or G.711
- Hear transcription, Speak voices, and a phone agent billed together, which AssemblyAI only bundles inside its Voice Agent API
AssemblyAI wins when transcription accuracy, diarization, and audio intelligence decide the project, and when you want those models and a voice agent from one vendor. PyAI wins when you want low per minute transcription cost, audio ready for phone lines, and text to speech on the same bill. Some teams run AssemblyAI for accuracy heavy work and PyAI for high volume or phone audio, since the per minute rate stays low as usage grows.
AssemblyAI cost by volume: hours of audio a month
Because AssemblyAI bills per hour with no plan, the monthly cost rises smoothly with volume. The first table maps each use case to what you need, the AssemblyAI model to pick, and what to watch for, and the second sets AssemblyAI against PyAI Hear at every volume.
| Use case | What you need | Model to pick | Watch out for |
|---|---|---|---|
| Podcasts and interviews | Up to about 50 hours | Universal-2 at $0.15/hr | Multichannel billed per channel |
| Meetings and notes | About 200 hours | Universal-2 or Pro | Diarization and add-ons extra |
| High volume or accuracy critical | 500 hours or more | Universal-3.5 Pro at $0.21/hr | Cost climbs with Pro and add-ons |
AssemblyAI vs PyAI monthly cost for every volume
AssemblyAI is shown at the Universal-2 base rate of $0.15 an hour. PyAI Hear is $0.001/min, about $0.06 an hour. Both scale linearly with no plan fee, so the gap holds steady as volume grows. Add-ons and multichannel would raise the AssemblyAI figure.
| Hours a month | AssemblyAI model | AssemblyAI cost | PyAI Hear |
|---|---|---|---|
| 10 | Universal-2 | $1.50 | $0.60 |
| 50 | Universal-2 | $7.50 | $3 |
| 100 | Universal-2 | $15 | $6 |
| 200 | Universal-2 | $30 | $12 |
| 500 | Universal-2 | $75 | $30 |
| 1,000 | Universal-2 | $150 | $60 |
AssemblyAI figures are Universal-2 pre-recorded at $0.15 an hour, before add-ons or multichannel. PyAI figures are Hear at $0.001/min, about $0.06 an hour, billed per minute with no plan. Universal-3.5 Pro at $0.21 an hour would be higher.
AssemblyAI pricing: the bottom line
AssemblyAI pricing is pay as you go with no contracts, from $0.15 an hour for Universal-2 and $0.21 for Universal-3.5 Pro pre-recorded, streaming from $0.15, and a flat $4.50 an hour Voice Agent API. A free tier gives $50 in credit, which AssemblyAI states covers about 185 hours, with no credit card. At 100 hours of pre-recorded audio a month, Universal-2 is $15. PyAI Hear is $6 for the same volume, billed per minute at $0.001/min.
PyAI is not a full product suite and does not match AssemblyAI on model accuracy, audio intelligence, or language coverage. PyAI is voice infrastructure that bills speech to text, text to speech, and a phone agent per minute on one invoice.
AssemblyAI pricing FAQs
The questions teams ask most before choosing AssemblyAI.
How much does AssemblyAI cost per hour?
AssemblyAI is pay as you go, billed per second of audio. Pre-recorded speech to text is $0.15 an hour with Universal-2 and $0.21 an hour with Universal-3.5 Pro. Streaming starts at $0.15 an hour, the Sync API is $0.45 an hour, and the all in one Voice Agent API is a flat $4.50 an hour.
Is AssemblyAI free?
AssemblyAI has a free tier with $50 in credit and no credit card required. AssemblyAI states the $50 credit covers about 185 hours of pre-recorded transcription or 333 hours of streaming. After the credit runs out you move to pay as you go, with no contracts, minimums, or monthly fee.
How much is AssemblyAI streaming and realtime transcription?
Streaming speech to text starts at $0.15 an hour on Universal-Streaming, the cost efficient live model. The highest accuracy live model, Universal-3.5 Pro Realtime, is $0.45 an hour. Streaming concurrency scales automatically with no extra fee, and the free tier includes about 333 hours of streaming.
What add-on fees does AssemblyAI charge?
Add-ons bill on top of the base rate per hour. Speaker diarization is about $0.02 an hour, keyterms prompting about $0.05 an hour, and medical mode about $0.15 an hour. Multichannel audio is billed per channel, so a one hour stereo file counts as two hours of transcription.
AssemblyAI vs Deepgram: which is cheaper?
AssemblyAI pre-recorded transcription is $0.15 to $0.21 an hour, while Deepgram Nova-3 pay as you go is $0.0043 a minute, about $0.26 an hour, per the Deepgram pricing page checked in November 2026. AssemblyAI is cheaper for batch audio, so price your real volume, streaming needs, and add-ons on each before switching.
What is a cheaper speech to text option than AssemblyAI?
For plain transcription by volume, PyAI Hear is $0.001/min, about $0.06 an hour, billed per minute with no contract and $50 of free credit, no credit card. AssemblyAI still leads on model accuracy, audio intelligence, and language coverage, so weigh accuracy and features against the per minute rate.
More AssemblyAI comparisons and guides
Price PyAI against your own volume.
Per minute pricing with no plan fee and no credits to burn. Speak is $0.04/min, Cast $0.02/min, Hear $0.001/min, and Omni $0.05/min + $0.01 telephony, with $50 in starter credit and no credit card.
