ElevenLabs API pricing: model rates and workload budgets

ElevenLabs API pricing at a glance

ElevenLabs text to speech (TTS) costs $0.08 per 1,000 input characters for v4, v3 and Multilingual v2, or $0.04 per 1,000 for v4 Turbo, v3 Conversational and Flash/Turbo. Scribe transcription starts at $0.22 per audio hour. Standalone Realtime API usage combines connected-audio minutes with input and output tokens.

Those are model rates. ElevenCreative credits and ElevenAgents hosted-call minutes have different allowances and billing rules. The API rate table prices usage in US dollars; all ElevenLabs amounts below exclude taxes and introductory discounts.

A $0.08 character-based speech rate and a $0.08 connected-call rate can produce different bills. They charge for different quantities.

TTS rates and monthly API allowances

TTS is metered by characters, not words or finished audio minutes. The two price groups are:

The rightmost column is gross model usage before a plan allowance. It is not an extra charge to add on top of a subscription that already covers some of those characters.

Regular month-to-month subscription prices and single-model TTS allowances are:

Enterprise terms are custom. First-month promotions and annual-prepayment equivalents are different from the recurring monthly prices above.

Free-plan audio has no commercial license and requires attribution when shared non-commercially. Paid-plan generation includes commercial rights subject to content rights, product terms and beta-service exclusions. The content licensing rules apply separately from usage allowances.

Treat each allowance as a single-product spending scenario, rather than adding every displayed model allowance together. A Creator subscription does not establish an entitlement to 275,000 v4 characters plus 550,000 v4 Turbo characters plus all its transcription hours.

Why audio-minute conversions distort a TTS budget

Characters per minute change with the script, language, pauses and delivery speed. A displayed approximate audio-minute equivalent does not change character-based metering.

For a workload budget, start with the text submitted for generation. Include additional successful generations used for pronunciation fixes, alternate deliveries and revised scripts. Do not infer how failed requests are charged from the character price alone.

Illustrative scenario assumption: Assume 10,000 responses average 80 characters, with another 25% of that text generated as alternate takes. The budget is 10,000 × 80 × 1.25 = 1,000,000 characters. At the unit rates, that is $80 for v4 or $40 for v4 Turbo before allowances.

The 25% is a planning input, not a measured retry rate. It makes the cost of discarded output explicit.

Speech-to-text API pricing

Speech to text (STT) charges for audio sent for transcription. Batch Scribe v2 and Scribe v2 Medical cost $0.22 per audio hour. Scribe v2 Realtime costs $0.39 per audio hour.

Optional keyterm prompting adds $0.05 per batch audio hour or $0.08 per realtime audio hour. Transcript editing adds $0.07 or $0.12 respectively and is an experimental feature.

Both batch and realtime transcription meter every second of submitted audio, including silence. There is no silence discount for STT. This differs from the hosted-agent silence rule. The transcription billing rules define the charged duration.

Illustrative scenario assumption: Assume 1,000 batch audio hours on Creator, no other allowance consumption and no optional features:

• Gross model usage: 1,000 × $0.22 = $220.
• Included usage: 100 hours.
• Additional usage: 900 × $0.22 = $198.
• Subscription plus additional usage consumed: $22 + $198 = $220.

Adding $220 to the subscription would count the included transcription twice. Enabling keyterm prompting also changes the workload price; the base-only calculation does not cover it.

Standalone Realtime API pricing

Eleven Realtime v1 is a speech-to-speech API. Its connected-audio charge is one component of the model bill:

Connected audio above the concurrency limit costs $0.16 per minute. The API table lists 4, 6, 10, 20, 30 and 40 concurrent calls across Free through Business.

Included-minute figures are estimates based on typical token usage. Creator lists approximately 262 Mini minutes or 216 Max minutes. Conversation length and tool use affect consumption, so these figures cannot support a precise remaining-minutes subtraction for an arbitrary token workload.

Illustrative scenario assumption: Assume 100 connected minutes, 1 million input tokens and 500,000 output tokens on Mini, with no burst:

100 × $0.08 + 1 × $0.20 + 0.5 × $6 = $11.20

This is a gross unit-rate calculation before any subscription allowance. It is neither a $0.08-per-minute total nor an ElevenAgents hosted-call estimate. Those two product paths have separate billing models.

Speech Engine and other audio APIs

The separate Speech Engine API lists $0.08 per additional minute and $0.16 per burst minute, with 275 minutes included on Creator. Its minute-based allowance is distinct from TTS characters and from the standalone Realtime API's audio-plus-token calculation.

Voice Isolator and Voice Changer each cost $0.12 per audio minute. Dubbing v1 costs $0.33 per source minute with a watermark or $0.50 without one; Dubbing Studio is $0.50 per source minute, and Dubbing v2 is $2.20 per source minute. These jobs belong in their own usage lines rather than in a TTS-character calculation.

Creative credits are a separate budgeting unit

ElevenCreative uses a shared monthly credit pool. Free includes 10,000 credits; Starter 30,000; Creator 121,000; Pro 600,000; Scale 1.8 million; and Business 6 million. Product, model and website-versus-API usage affect credit consumption.

That is why Creator's 121,000 Creative credits and 275,000 v4 API characters are different figures. A single credit-to-character conversion cannot price a mixed workload of narration, transcription, dubbing and other creation tools. The Creative credit rules describe the shared pool.

On an active paid subscription, up to two unused monthly credit allotments can roll over. Free credits do not roll over. Cancellation or downgrade forfeits unused paid subscription credits at the end of the billing cycle. Prepaid pay-as-you-go funds follow separate expiry rules.

Pay-as-you-go funding is different from an overage invoice

For new self-serve accounts, pay as you go (PAYG) is prepaid. Included subscription quota is consumed first, followed by prepaid funds. A paid subscription is not required to add PAYG funds.

The minimum top-up is $5. Funds expire 12 months after purchase. The wallet keeps a dollar value, although other parts of the platform display its credit equivalent at the active plan's rate. Buying funds does not itself upgrade plan features.

If the balance reaches zero with auto top-up off or failing, service pauses. Auto top-up can charge a saved payment method when its threshold is reached. Without two-factor authentication, monthly purchase caps are $100 on Free, Starter and Creator, $200 on Pro, $600 on Scale and $2,000 on Business. These mechanics are defined in the PAYG billing documentation.

Enterprise and eligible older subscriptions can use postpaid usage-based billing instead. Their account-specific overage terms are not interchangeable with new self-serve PAYG.

Separate consumption from cash paid

For one million v4 Turbo characters on Creator:

• The $22 subscription includes 550,000 characters.
• The remaining 450,000 consume $18 at $0.04 per 1,000.
• Subscription plus additional usage consumed is $40.

If the wallet starts empty and you purchase $20 of funds, cash paid is $42 including the subscription. The workload leaves $2 in the wallet. That balance is prepaid capacity, not another $2 of usage consumed.

This distinction matters when finance reconciles top-up payments against a month's actual consumption.

ElevenAgents hosted-call pricing

ElevenAgents bills hosted conversations by call minutes. Its hosted-agent pricing includes these allowances:

Additional standard minutes cost $0.08. When burst capacity is enabled, calls above normal concurrency cost $0.16 per minute. Text messages cost $0.003 each. Text-only agents incur message and LLM charges; multimodal agents can incur both call-minute and message charges.

Voice and multimodal calls use connection duration, which can exceed speaking time. Periods of silence longer than ten seconds receive a 95% discount on the affected hosted-call charge. That discount does not apply to a separate carrier bill or to standalone STT. The call-duration billing rules explain the clock and channel differences.

For a 2,000-minute Creator scenario, 275 minutes are included and 1,725 additional standard minutes cost $138. The $160 subtotal includes the $22 subscription. Assume no qualifying silence discount and no burst. Replacing ten of those additional standard minutes with ten burst minutes adds $0.80.

LLM usage is extra. ElevenLabs adds no separate telephony fee, but an external phone provider bills its own carrier charges. A complete hosted-call budget therefore has separate lines for:

1. Subscription and additional hosted-call usage.
2. Burst capacity and messages where used.
3. LLM input, output and other chargeable model usage.
4. External carrier minutes, phone numbers and other provider fees.
5. Taxes and any additional contracted services.

For a custom stack using separate TTS and STT APIs, budget those speech components and your orchestration costs instead. Do not automatically add standalone speech-API charges to an ElevenAgents plan that already includes speech processing.

Reproduce a single-product monthly budget

Use the same formula for a homogeneous workload:

subscription + max(0, quantity − included quantity) × unit rate

Quantities and rates must use matching units. For a per-1,000-character rate, divide excess characters by 1,000. For an hourly transcription rate, use audio hours.

The following Python uses only the standard library and makes no API requests. Each row is an independent Creator-plan scenario, with no starting rollover, other quota consumption, optional features or tax:

from decimal import Decimal as D


def monthly_budget(quantity, included, rate, unit_size=1):
    excess = max(D("0"), D(str(quantity)) - D(str(included)))
    additional_usage = excess / D(str(unit_size)) * D(rate)
    return D("22") + additional_usage


scenarios = [
    ("v4: 1M characters", 1_000_000, 275_000, "0.08", 1_000),
    ("v4 Turbo: 1M characters", 1_000_000, 550_000, "0.04", 1_000),
    ("Scribe v2: 1,000 hours", 1_000, 100, "0.22", 1),
    ("ElevenAgents: 2,000 minutes", 2_000, 275, "0.08", 1),
]

for name, quantity, included, rate, unit_size in scenarios:
    cost = monthly_budget(quantity, included, rate, unit_size)
    print(f"{name}: ${cost:.2f}")

Output:

v4: 1M characters: $80.00
v4 Turbo: 1M characters: $40.00
Scribe v2: 1,000 hours: $220.00
ElevenAgents: 2,000 minutes: $160.00

These are subscription-plus-additional-consumption estimates, not guaranteed cash invoices. The agent row also excludes LLM usage, carrier charges, messages and burst minutes. The formula does not allocate a shared allowance across several products or estimate Realtime token consumption.

When the budget is for a production voice agent

A standalone speech API fits teams that want to build and operate the surrounding application themselves. A production agent also needs conversation control, tools, telephony integration, testing and monitoring. Those responsibilities belong in the architecture decision alongside model spend.

Our managed voice AI runtime serves technical teams building complete conversational AI products. We provide the runtime and operations around the conversation rather than a standalone narration or voice-cloning studio.

Our Dasha pricing plans include a free Developer offer with 1,000 free minutes and one concurrent call. Growth starts at $0.08 per minute, with one-second billing and no minimum call duration. Voice over IP (VoIP) and LLM-token costs are separate. This is a platform pricing component, not a like-for-like TTS character rate or a total call bill.

If your project needs a managed production runtime, start building with Dasha using our free Developer offer.