AssemblyAI Pricing Explained (2026): Streaming, Voice Agent API and What a Month Really Costs
AssemblyAI is pay as you go, priced per hour of audio, with no minimums, contracts or upfront fees, and $50 of free credit to start. Streaming, pre-recorded transcription and its all-in-one Voice Agent API are priced separately, and optional features add a few cents an hour each. Every figure below is from assemblyai.com/pricing and AssemblyAI's published pricing reference, read October 22, 2026.
AssemblyAI bills streaming for as long as the connection is open, not just while people talk. Close the socket the moment a call ends and your bill will match your call log.
AssemblyAI streaming starts at $0.15 an hour, billed on how long the connection is open. Here's every AssemblyAI price line (Universal-Streaming, Universal-3.6 Pro Realtime, add-ons, pre-recorded and the $4.50/hr Voice Agent API), what 1,000 to 100,000 minutes a month cost, and the billing rules that change the total.
What you pay for on every AssemblyAI minute
Streaming speech-to-text
$0.0025–$0.0075/min
Universal-Streaming (English only) and Universal-Streaming Multilingual (English, Spanish, German, French, Portuguese and Italian) are $0.15 an hour. Universal-3.6 Pro Realtime, AssemblyAI's most accurate real-time model for voice agents (32 languages, keyterm prompting included), is $0.45 an hour. Billed on how long the WebSocket is open, not on audio sent.
Optional add-ons
+$0.04–$0.15/hr each
Only if you switch them on, and charged for the whole session: speaker diarization $0.12, Medical Mode $0.15, PII text redaction $0.12, prompting $0.05 and Voice Focus (isolates the main speaker) $0.10 on Universal-3.6 Pro Realtime. Keyterm prompting is $0.04 on Universal-Streaming and included on Universal-3.6 Pro.
Phone line
Not from AssemblyAI
Speech-to-text doesn't include telephony; your carrier or voice platform bills the call. With the Voice Agent API you connect your own Twilio account, and AssemblyAI says it adds no markup to your carrier rate.
Total per minute
about 0.3¢ to 0.8¢
Streaming transcription only, pay as you go, before add-ons. The Voice Agent API ($0.075 a minute, speech-to-text, LLM and voice included) is a separate product.
What a month costs on AssemblyAI’s own numbers
1,000 min/month
- Universal-Streaming
- $2.50
- 3.6 Pro Realtime
- $7.50
- 3.6 Pro + Medical
- $10.00
- Voice Agent API
- $75.00
5,000 min/month
- Universal-Streaming
- $12.50
- 3.6 Pro Realtime
- $37.50
- 3.6 Pro + Medical
- $50.00
- Voice Agent API
- $375.00
10,000 min/month
- Universal-Streaming
- $25.00
- 3.6 Pro Realtime
- $75.00
- 3.6 Pro + Medical
- $100.00
- Voice Agent API
- $750.00
25,000 min/month
- Universal-Streaming
- $62.50
- 3.6 Pro Realtime
- $187.50
- 3.6 Pro + Medical
- $250.00
- Voice Agent API
- $1,875.00
50,000 min/month
- Universal-Streaming
- $125.00
- 3.6 Pro Realtime
- $375.00
- 3.6 Pro + Medical
- $500.00
- Voice Agent API
- $3,750.00
100,000 min/month
- Universal-Streaming
- $250.00
- 3.6 Pro Realtime
- $750.00
- 3.6 Pro + Medical
- $1,000.00
- Voice Agent API
- $7,500.00
AssemblyAI's pay-as-you-go rates times minutes of open connection, with no platform fee: Universal-Streaming $0.15 an hour, Universal-3.6 Pro Realtime $0.45, the same with Medical Mode $0.60, and the Voice Agent API $4.50. The first three columns are transcription only (your LLM, voice and phone line are billed by their vendors); the Voice Agent API column includes speech-to-text, the LLM and text-to-speech but not the carrier. Idle connection time counts, multichannel audio is billed per channel, and new accounts spend the $50 free credit first. Want your own volume and carrier? Try the cost calculator.
Packages and add-ons
Free start
$50 free credit
No credit card required. AssemblyAI's FAQ says the free tier covers up to 185 hours of pre-recorded or up to 333 hours of streaming transcription. Free accounts can open 5 new streaming connections a minute.
Pay as you go
No minimums
Billed monthly on actual usage, invoiced at the start of each month for the month before, with no commitments, upfront fees or contracts. Streaming starts at 100 new sessions a minute and scales up automatically by 10% whenever you use 70% of your limit, with no ceiling and no concurrency fees; higher starting limits are free on request.
Pre-recorded transcription
$0.15–$0.21/hr
Universal-3.5 Pro $0.21 an hour (18 languages, AssemblyAI's most accurate file model) and Universal-2 $0.15 an hour (99 languages). Add-ons: speaker diarization $0.02, prompting $0.05, keyterm prompting $0.05 and Medical Mode $0.15 an hour. Multichannel files are billed per channel, so a 1-hour stereo file counts as 2 hours.
Sync and Dictation APIs
$0.30 and $0.62/hr
The Sync API returns Universal-3.5 Pro transcripts for clips up to 2 minutes in one call (AssemblyAI quotes about 134 ms p50) at $0.30 an hour. The Dictation API turns a single spoken clip of up to 120 seconds into finished text at $0.62 an hour, all features included.
Voice Agent API
$4.50/hr ($0.075/min)
Universal-3.6 Pro Realtime speech-to-text with prompting and Voice Focus, turn and interruption detection, AssemblyAI's Voice Agent LLM, its text-to-speech voices and hosting over one WebSocket, plus recordings and transcripts. Billed per second on connected conversation time, with no per-layer add-ons, concurrency fees or per-agent subscriptions. Connect your own Twilio account for phone calls.
Speech Understanding and Guardrails
+$0.01–$0.15/hr
On pre-recorded audio: speaker identification $0.02 ($0.10 on medium effort), translation $0.06, custom formatting $0.03, entity detection $0.08, sentiment analysis $0.02, key phrases $0.01, topic detection $0.15 and summarization $0.02 ($0.07 medium). Guardrails: profanity filtering $0.01, PII audio redaction $0.05, PII text redaction $0.08 and content moderation $0.15.
LLM Gateway
Per 1M tokens
One API key for LLMs from several providers, billed per million input and output tokens at each model's rate, for example GPT-5 Nano $0.05 / $0.40, Gemini 2.5 Flash $0.30 / $2.50, Sonnet 4.6 $3.00 / $15.00 and GPT-5.5 $5.00 / $30.00.
Volume and enterprise
Contact sales
Custom pricing, tiered volume discounts and enterprise agreements for high usage, plus custom rate limits. AssemblyAI is also sold through the AWS Marketplace. Its pricing reference says the EU region costs the same as the US and a HIPAA BAA can be signed self-serve at no extra cost.
How it compares on published pricing
AssemblyAI- Starting price
- Streaming from $0.15/hr; $50 free credits
- Best for
- Low-cost streaming and batch transcription
Deepgram- Starting price
- Nova-3 streaming $0.0048/min ($0.29/hr); $200 free credit
- Best for
- Streaming speech-to-text for voice agents
Speechmatics- Starting price
- From $0.24/hr pay as you go (as low as ~$0.12/hr with discounts)
- Best for
- Multilingual transcription with language switching
Soniox- Starting price
- ~$0.10/hr async; ~$0.12/hr real-time
- Best for
- Very low-cost speech-to-text
Gladia
- Starting price
- Async from €0.16/hr; real-time €0.68/hr
- Best for
- Real-time multilingual transcription
STT models compared
- Price
- Free until Nov 19, 2026; then $0.002/min
- Latency
- —
- Languages
- —
- Price
- $0.0065/min (English)
- Latency
- Built-in turn detection
- Languages
- English or multilingual
- Price
- $0.39 / hr
- Latency
- ~150 ms partials
- Languages
- 90+
- Price
- $0.0048/min streaming (monolingual, promo; reg. $0.0077)
- Latency
- —
- Languages
- 45+ (Nova models)
- Price
- $0.21 / hr
- Latency
- Async (pre-recorded)
- Languages
- 18
- Price
- $0.017/min
- Latency
- Live transcription
- Languages
- —
- Price
- $0.30 / hr real-time
- Latency
- Built-in turn detection and VAD (real-time)
- Languages
- 55+
- Price
- $0.40 / hr real-time; $0.24 / hr files
- Latency
- Real-time or files
- Languages
- 56, switches mid-sentence
- Price
- $0.30 / hr files
- Latency
- Files (real-time coming soon)
- Languages
- 55+, tuned for clinical terms
- Price
- ~$0.12 / hr real-time; ~$0.10 / hr async
- Latency
- Real-time streaming
- Languages
- 60+ (translation included)
- Price
- €0.16 / hr async (real-time €0.68 / hr)
- Latency
- Async (files)
- Languages
- 100+
See all voice models · Latency and prices as each vendor publishes them.
What to check before you choose
- Close the connection when the call ends. Streaming is billed on how long the WebSocket is open, not on how much audio you send, so idle time counts: AssemblyAI's own pricing reference says a socket left open for 5 minutes between calls is 5 billable minutes.
- Pick the model for the job. Universal-Streaming is $0.15 an hour and Universal-3.6 Pro Realtime is $0.45. At 10,000 minutes a month that's $25 against $75. AssemblyAI itself recommends Universal-3.6 Pro (or the Voice Agent API) for voice agents, and keyterm prompting is included on it.
- Add-ons stack on every session minute. Universal-3.6 Pro with speaker diarization ($0.12), Medical Mode ($0.15), Voice Focus ($0.10) and prompting ($0.05) comes to $0.87 an hour, almost double the base rate. Turn on only what your calls need.
- The Voice Agent API is the bundled option. At $0.075 a minute it covers listening, the LLM and the voice, the same as Deepgram's Voice Agent Standard tier ($0.075). ElevenLabs Agents is $0.08 a minute plus the LLM, and Vapi's calculator puts its default stack at about 8.2 to 12.9¢ before carrier fees. With AssemblyAI the phone call is still billed by your Twilio account (US local inbound is $0.0085 a minute).
- Count channels on recorded calls. Multichannel audio is billed per channel, so a 1-hour stereo call recording is 2 billable hours. Downmix if you don't need a transcript per side.
- Compare per hour against other speech APIs. AssemblyAI streaming starts at $0.15 an hour. On their own pricing pages, re-checked today, Deepgram Nova-3 streaming is $0.29 an hour at a limited-time promotional rate ($0.46 regular), Soniox real-time is about $0.12, Speechmatics real-time is $0.30 and Gladia real-time is €0.68 an hour. For trying them out, AssemblyAI's $50 free credit compares with $200 at Deepgram, $100 at Speechmatics and €50 at Gladia.
- Free accounts are rate-limited. The free tier opens 5 new streams a minute; pay as you go starts at 100 a minute and grows automatically. Load-test on a paid account before launch.
Questions people ask
How much does AssemblyAI cost per hour?
Streaming speech-to-text is $0.15 an hour on Universal-Streaming (English or multilingual) and $0.45 an hour on Universal-3.6 Pro Realtime. Pre-recorded transcription is $0.15 an hour on Universal-2 and $0.21 on Universal-3.5 Pro. The Voice Agent API, with speech-to-text, LLM and voice included, is $4.50 an hour ($0.075 a minute). New accounts get $50 in free credit.
How is AssemblyAI streaming billed?
On session time: how long the WebSocket connection stays open, not the length of audio you send. Idle time counts, so close the connection as soon as a call ends. Optional streaming add-ons such as diarization are charged for the whole session too.
Is AssemblyAI free to start?
Yes. You can sign up and start transcribing with no credit card, and new accounts get $50 in free credit. AssemblyAI's FAQ says that covers up to 185 hours of pre-recorded or up to 333 hours of streaming transcription. Free accounts can open 5 new streaming connections a minute.
What does the AssemblyAI Voice Agent API cost?
$4.50 an hour, or $0.075 a minute, billed per second on connected conversation time. It includes Universal-3.6 Pro Realtime speech-to-text, turn and interruption detection, AssemblyAI's Voice Agent LLM, its text-to-speech voices, hosting, and recordings and transcripts. There are no per-layer add-ons, concurrency fees or per-agent subscriptions. Phone calls come through your own Twilio account, which bills the carrier minutes.
What do AssemblyAI add-ons cost?
On streaming: speaker diarization $0.12 an hour, Medical Mode $0.15, PII text redaction $0.12, prompting $0.05 and Voice Focus $0.10 (Universal-3.6 Pro only), and keyterm prompting $0.04 on Universal-Streaming (included on Universal-3.6 Pro). On pre-recorded audio, diarization is $0.02 an hour and Medical Mode $0.15. Add-ons stack on the base rate.
Does AssemblyAI limit concurrent streams?
Not with a fixed cap. Pay-as-you-go accounts start at 100 new streaming sessions a minute, and the limit rises 10% automatically whenever you use 70% of it, with no ceiling and no concurrency fees. Free accounts can open 5 new streams a minute.
Is AssemblyAI HIPAA compliant?
AssemblyAI's pricing reference says a HIPAA Business Associate Agreement can be signed self-serve, without a sales call, at no extra cost, and lists SOC 2 Type 2, ISO 27001 and GDPR, with the Voice Agent API PCI-certified. It also offers an EU region at the same price as the US.
Ready to look closer? See AssemblyAI’s full listing and price history →
Not sure which fits? Answer a few quick questions in the finder →
Prices from vendors' official pricing pages, checked October 22, 2026. Not sponsored; order isn't a ranking.