Short answer. ElevenLabs Speech to Text now costs $0.20 per hour of audio for batch transcription and $0.22 per hour for realtime transcription, for every Scribe model on all self-serve and Business plans, according to an entry dated November 17, 2026 in the ElevenLabs changelog. The changelog says Enterprise pricing is unchanged. In a hypothetical month of 20,000 call minutes transcribed in real time, that is $73.33, against $130.00 at the $0.39 an hour we recorded for Scribe v2 Realtime in September 2026.

This guide covers the rate for each Scribe model, the add-ons and the hours each plan includes, how ElevenLabs meters audio, a worked monthly bill for a calling team, and how the new rates compare with OpenAI and Deepgram. Prices exclude taxes, as the page states, and were read on ElevenLabs's API pricing page on November 19, 2026.
Key numbers
| Item | Number |
|---|---|
| Batch, every Scribe model, per hour (self-serve and Business plans) | $0.20 |
| Realtime, every Scribe model, per hour (self-serve and Business plans) | $0.22 |
| Changelog entry dated | November 17, 2026 |
| Realtime rate we recorded in September 2026 (the new rate is 43.6% lower, our arithmetic) | $0.39 an hour |
| Keyterm prompting, batch and realtime, per hour | $0.050 and $0.080 |
| Transcript editing, batch and realtime, per hour | $0.070 and $0.120 |
| Pro plan ($99) hours included | 495 batch, 450 realtime |
| Maximum file | 5 GB, 15 hours |
| OpenAI gpt-transcribe, per minute ($0.27 an hour) | $0.0045 |
| Worked example, 20,000 minutes realtime (hypothetical) | $73.33 a month |
Rates read on 19 November 2026 from ElevenLabs's API pricing page, changelog and Speech to Text documentation, and from OpenAI's pricing page; the September rate is from our articles of 7 and 16 September 2026; comparisons and the worked example are our arithmetic.
What changed in the November 17 changelog

The changelog entry reads: "Speech to Text now costs $0.20 per hour of audio for batch transcription and $0.22 per hour for realtime transcription, for every Scribe model on all self-serve and Business plans. Enterprise pricing is unchanged." It does not print the old rates, so the "before" column below is what our own earlier articles recorded from ElevenLabs's pricing page, not a figure from the changelog.
| Mode | Last rate we recorded | Rate in the entry dated November 17, 2026 | Change |
|---|---|---|---|
| Realtime (Scribe v2 Realtime) | $0.39 an hour (our articles of September 16 and October 2, 2026) | $0.22 an hour | 43.6% lower |
| Batch (Scribe v2) | $0.22 an hour (our article of September 7, 2026) | $0.20 an hour | 9.1% lower |
Against our last readings, the realtime drop is the larger one, and realtime transcribes audio as it is spoken, which is what a phone agent needs during a call. ElevenLabs may have changed its rates between our readings and November 17, so treat the percentages as our arithmetic against the last rates we saw, not as ElevenLabs's statement of its old prices.
Scribe prices per hour by model
On self-serve and Business plans, every Scribe model now sits on one of two rates. The Speech to Text documentation says to "Use Scribe v3 for recorded audio and Scribe v3 Live for real-time audio." Scribe v3 and Scribe v3 Live came out in the changelog entry of November 9, 2026, which says they "succeed Scribe v2 and Scribe v2 Realtime, which remain available."
| Model | Mode | Price per hour | Per minute (our arithmetic) |
|---|---|---|---|
| Scribe v3 | Batch (recorded audio) | $0.20 | $0.0033 |
| Scribe v2 | Batch | $0.20 | $0.0033 |
| Scribe v2 Medical | Batch, clinical audio | $0.20 | $0.0033 |
| Scribe v3 Live | Realtime | $0.22 | $0.0037 |
| Scribe v2 Realtime | Realtime | $0.22 | $0.0037 |
Because the old and new generations now cost the same, the choice between them is about features, not price. The documentation lists speaker diarization for up to 32 speakers and keyterm prompting for up to 1,000 terms on Scribe v3, and turn prediction plus a keyterm library of up to 10,000 terms on Scribe v3 Live. ElevenLabs reports that Scribe v3 makes 27% fewer word errors than Scribe v2 on the FLEURS benchmark; that is a figure ElevenLabs reports, so we would check accuracy on a sample of your own calls before switching models. Scribe v2 Medical is listed as HIPAA-eligible, with BAAs and Zero Retention Mode.
Add-ons and hours included by plan

The pricing page lists the same hourly rate in every column of its pricing table, from Free / pay as you go up to Business. Add-ons are billed on top of the base rate (two on batch, three on realtime), and each column lists the hours included.
| Plan (monthly price) | Batch hours included | Realtime hours included |
|---|---|---|
| Free / pay as you go | 4 hours 30 minutes | 2 hours 30 minutes |
| Starter ($6, first month $1) | 30 hours | 27 hours |
| Creator ($22, first month $11) | 110 hours | 100 hours |
| Pro ($99) | 495 hours | 450 hours |
| Scale ($299) | 1,495 hours | 1,359 hours |
| Business ($990) | 4,950 hours | 4,500 hours |
For each paid plan, the included hours roughly equal the plan price divided by the hourly rate, rounded down on the page: $99 divided by $0.20 is 495 hours, and $99 divided by $0.22 is 450. The add-ons, per hour of audio: keyterm prompting costs $0.050 on batch and $0.080 on realtime, the keyterm library (Scribe v3 Live only) costs $0.080, and transcript editing costs $0.070 on batch and $0.120 on realtime. Keyterm prompting lets you supply terms such as product and prospect names (up to 1,000 terms on batch, 50 on realtime, per the documentation); we would test whether it fixes them on your own calls. On batch it adds a quarter to the base rate ($0.05 on $0.20).
How ElevenLabs bills transcription

ElevenLabs bills "based on the duration of the audio sent for transcription", per hour of audio. The documentation adds: "Silence counts toward the billed duration. Both batch and realtime transcription are metered on every second of audio you send, including pauses and silent stretches, and there is no discount for silence." For call recordings with hold music or long pauses, trimming silence and hold music before upload shortens the billed duration.
- File limits: files up to 5 GB, and up to 15 hours of audio in standard mode. In multichannel mode the combined duration of all channels must be under 15 hours.
- Multichannel: up to 5 channels, "each channel is processed independently and assigned a speaker ID based on its channel number". This suits stereo call recordings with the rep on one channel and the prospect on the other. The documentation does not say whether a two-channel file bills once or per channel; Deepgram's page, by contrast, says its "Multichannel audio bills by total processed duration across channels".
- Speed on long files: batch files over 8 minutes are split and transcribed in parallel, with a concurrency of up to 4. A 15 minute file runs at a concurrency of 2 and a 120 minute file at 4, by the documentation's own examples.
What a calling team pays per month

Take a hypothetical team whose AI phone agent handles 20,000 call minutes a month (333.33 hours of audio), with realtime transcription during each call and a batch pass over the recordings afterwards for quality review. It assumes single-channel audio billed on clock time and transcription bought through the API. All figures are our arithmetic on the current published rates and on the rates our articles recorded in September 2026.
| Line | Rate per hour | Monthly cost |
|---|---|---|
| Realtime, new rate | $0.22 | $73.33 |
| Realtime, at the $0.39 we recorded in September | $0.39 | $130.00 |
| Realtime with keyterm prompting | $0.30 | $100.00 |
| Batch review of the recordings, new rate | $0.20 | $66.67 |
| Batch with keyterm prompting | $0.25 | $83.33 |
| Realtime plus batch, new rates | $0.42 | $140.00 |
| Realtime plus batch, at our September readings | $0.61 | $203.33 |
Against the rates we recorded in September, this workload costs $56.67 less a month on the realtime leg and $63.33 less on both legs together. Transcription is a small share of an AI call: our AI phone call cost per minute breakdown calls speech to text "the cheapest leg in the stack". Still, a lower transcription rate makes a second batch pass for coaching notes and CRM summaries cheaper to justify.
How it compares with OpenAI and Deepgram

| Service | Published rate | Per hour (our arithmetic) | 20,000 minutes |
|---|---|---|---|
| ElevenLabs Scribe, batch | $0.20 an hour | $0.20 | $66.67 |
| ElevenLabs Scribe, realtime | $0.22 an hour | $0.22 | $73.33 |
| OpenAI gpt-transcribe | $0.0045 a minute | $0.27 | $90.00 |
| OpenAI gpt-live-transcribe | $0.017 a minute | $1.02 | $340.00 |
| Deepgram Nova-3 Monolingual, pre-recorded | $0.0043 a minute | $0.258 | $86.00 |
| Deepgram Nova-3 Monolingual, streaming | $0.0048 a minute (was $0.0077) | $0.288 | $96.00 |
Rates were read on November 19, 2026 from OpenAI's pricing page and Deepgram's pricing page; Deepgram rates are its Pay as you go tier, and the page labels its streaming rates as promo pricing. On these published rates, ElevenLabs is now the cheapest of the three for both recorded and live audio, and its realtime rate is under a quarter of gpt-live-transcribe's. Published rates are not the whole bill: Deepgram bills per second with no block-based rounding for normal audio requests (ElevenLabs also meters every second), and its Growth tier lowers the rates for teams that prepay from $4,000 a year. For the wider field, our AI transcription cost per hour comparison covers more vendors from September 2026, and our Whisper API pricing guide covers OpenAI's deprecated transcription models.
What to check before you switch
These are our suggestions, not ElevenLabs's requirements.
- Test on your own calls. Run 50 to 100 recorded calls through the model you plan to use and check names, numbers and the words your team actually says. Benchmark figures here are vendor-reported.
- Count the add-ons. Keyterm prompting changes the realtime rate from $0.22 to $0.30 an hour, so price the setup you will run, not the base rate.
- Ask how stereo files bill. If your recordings are two-channel, confirm with ElevenLabs whether you pay for one hour or two.
- Check consent before you transcribe. Recording a call comes first, and the rules can differ by state; see our call recording laws by state guide.
- Enterprise contracts are separate. The changelog says Enterprise pricing is unchanged, so a team on a negotiated contract should ask its account manager rather than assume the new rate applies.
If you run an AI voice agent, the agent platform's bundled per-minute price may already include speech recognition. Our ElevenLabs Agents pricing guide and our list of the best AI voice agents in 2026 show how those bundles are priced, and our AI answering service for HVAC guide shows one trade's use case. The changelog does not say whether ElevenAgents' own prices moved with this cut.
For AI calling setups, see our AI calling for home services page and our AI automation agency services. For a broader view of what AI tools cost a small team, see our AI chatbot for business cost guide.
Frequently Asked Questions
From the changelog entry dated November 17, 2026, it costs $0.20 per hour of audio for batch transcription and $0.22 per hour for realtime, for every Scribe model on all self-serve and Business plans. Enterprise pricing is unchanged.
Not beyond a small allowance. The pricing page lists 4 hours 30 minutes of batch and 2 hours 30 minutes of realtime as hours included in its Free / pay as you go column, read on November 19, 2026. Every column lists $0.20 an hour for batch and $0.22 an hour for realtime.
Scribe v3 transcribes recorded audio and Scribe v3 Live transcribes in real time. Scribe v3 costs $0.20 an hour and supports diarization of up to 32 speakers; Scribe v3 Live costs $0.22 an hour and offers turn prediction, multilingual mode and a keyterm library of up to 10,000 terms.
Yes. The documentation says both batch and realtime transcription are metered on every second of audio you send, including pauses and silent stretches, with no discount for silence.
On list price, yes. ElevenLabs batch is $0.20 an hour against $0.27 an hour for OpenAI's gpt-transcribe ($0.0045 a minute), and ElevenLabs realtime is $0.22 an hour against $1.02 for gpt-live-transcribe ($0.017 a minute), read on November 19, 2026.
No. The changelog entry dated November 17, 2026 says Enterprise pricing is unchanged; the new rates apply to self-serve and Business plans.