ElevenLabs API Pricing vs App Plans: Buying Guide

Short answer: budget ElevenCreative work, including Studio, separately from ElevenAPI traffic, but do not assume ElevenLabs requires two subscriptions. Most API endpoints are available across plans, including Free, and subscription usage can draw from the same account or workspace credit pool.

The API rate card nevertheless measures work in its native units—characters for text to speech, audio time for speech to text, call minutes for Speech Engine, and source minutes for dubbing—and offers Pay As You Go. Treat the routes as two forecast lines that may settle against one shared plan and an optional prepaid balance.

Pick the route by who does the work

ElevenCreative, including Studio, is the creator-led route. A person imports or writes content, chooses voices, edits timing, adds music or effects, reviews generations, collaborates, and exports a finished asset. The subscription decision therefore includes workflow entitlements as well as usage: commercial rights, project allowance, cloning tier, output quality, collaboration, and seats can matter before the raw credit total does.

ElevenAPI is the software-led route. An application authenticates with an API key and generates, transcribes, processes, or dubs content programmatically. Its budget starts with the endpoint and model, then multiplies the official unit rate by expected traffic. Retries, test environments, batch jobs, peak concurrency, and unplanned user demand belong in that forecast even when the resulting usage draws from the same subscription pool.

This is a budgeting split, not a claim that the vendor always issues two bills. One ElevenLabs subscription can support app and API work, while usage analytics can separate consumption by product and API key.

Use app plans for creator-led production

The app route makes most sense when the deliverable is produced inside ElevenCreative Studio. Studio combines narration, video, captions, music, and sound effects on a timeline, supports review and commenting, and exports audio or video. That editing surface is the value being purchased; an API-only estimate does not account for the time saved by project organization, sentence-level revision, or human review.

Plan gates change the buying decision. Free provides a small monthly credit allowance and a limited number of Studio projects for evaluation. Starter adds a commercial license, Instant Voice Cloning, more Studio projects, and Dubbing Studio. Creator adds Professional Voice Cloning. Pro and higher plans add higher-quality output options, while Scale and Business move the purchase toward collaboration and workspace ownership.

Creator usage is drawn from a shared monthly credit pool across products. Text generation, transcription, music, effects, voice processing, and dubbing do not each receive an isolated allowance. The pricing page also says credits are charged per generation request rather than per download, with only limited qualifying regenerations available at no charge. A realistic creator forecast therefore needs a revision margin instead of equating finished audio minutes with total consumption.

Forecast API usage in native units

ElevenLabs says most API endpoints are available on all plans, including Free, and an API request deducts usage from the account just as website generation does. At the same time, the ElevenAPI rate card presents API consumption in US-dollar rates and shows how much model-specific usage is included at each subscription tier. Those statements fit together: the subscription can fund API calls, but the API workload still needs its own meter-based forecast.

For text to speech, the current rate card lists Flash and Turbo at $0.05 per 1,000 characters and Multilingual v2 or v3 at $0.10 per 1,000 characters. Speech to text uses audio time instead: Scribe v2 is listed at $0.22 per hour and Scribe v2 Realtime at $0.39 per hour. Do not turn these into one universal “cost per finished minute.” Script density, language, model choice, silence, and retries can change the relationship between characters, input time, and usable output.

Real-time products need another line. Speech Engine is priced by call minute, and the plan table separately exposes included minutes, additional-call pricing, burst pricing, and concurrency. A production forecast should use expected connected minutes plus peak simultaneous calls, then keep a reserve for bursts and failed or repeated sessions. Realtime transcription should remain in its own audio-hour estimate rather than being mixed into text-to-speech characters.

Pay As You Go is an extension path, not a feature-tier replacement. ElevenLabs documents that subscription credits are consumed first and a PAYG balance is used after them. A PAYG balance can coexist with Free or a paid subscription, but voice slots, access tiers, concurrency, and other plan limits still come from the underlying subscription. Use top-ups for variable consumption; upgrade the plan when an entitlement or operating limit is the blocker.

Separate cloning and dubbing decisions

Voice cloning has both an entitlement boundary and a usage boundary. Starter unlocks Instant Voice Cloning, while Professional Voice Cloning requires Creator or above. Instant cloning uses short reference audio and is available quickly; professional cloning trains a dedicated model from substantially more audio and takes longer. The plan unlocks the creation workflow, while subsequent synthesis still consumes app or API usage.

Custom voices do not have to stay trapped in the app. ElevenLabs says voices in a user's library—including professional, cloned, and designed voices—can be referenced by ID through the API. That makes a mixed route legitimate: create and govern the voice in the workspace, then invoke it from software. The person creating an Instant Voice Clone must confirm that they have the right and consent to clone it, so authorization remains a separate requirement from paying for a tier.

Dubbing has a similar split. Dubbing Studio is for granular transcript edits, speaker reassignment, and per-clip regeneration inside a human review workflow. The Dubbing API is measured by source-audio minute and is better suited to repeatable automation. A team that prototypes and corrects a dub in Studio but later runs localization through the API should track both workloads, because each can consume the shared pool even though the operational jobs are different.

Move to workspace or enterprise when governance is the constraint

A team route becomes relevant when shared assets, centralized billing, member access, and durable API ownership matter. ElevenLabs workspaces distinguish Full Seats from Basic Seats, and multi-seat customers can use service-account API keys that are not tied to one employee. That is safer for backend systems than building production around a personal key whose access changes when the user leaves.

Scale and Business are the self-serve collaboration steps on the pricing page. Enterprise is the sales-led route for custom seats and credits, SSO, DPA or SLA terms, BAAs for eligible healthcare customers, elevated concurrency, managed dubbing, discounts at scale, and priority support. The API pricing page also calls out dedicated support, custom rate limits, SOC 2 compliance, and master service agreements.

Contact sales when security review, contractual assurance, rate limits, regional or compliance requirements, managed delivery, or unusually large volume determines whether the system can launch. If the only problem is a variable pilot workload, a controlled PAYG balance may be the narrower answer.

Build the budget before you buy

Create two usage lines even if they share one account:

1. For the creator line, estimate active editors, Studio projects, cloning method, finished media, expected revisions, dubbing review, export quality, and commercial-use requirements.
2. For the API line, estimate each model or endpoint separately in characters, audio hours, call minutes, or source minutes. Add development traffic, retries, peak concurrency, and a production reserve.
3. Reconcile both lines against the subscription's shared credit pool. If the combined forecast can exceed it, decide whether a higher tier or a capped PAYG balance is the safer overflow route.
4. Check ownership last: personal versus service-account keys, workspace permissions, seat types, procurement terms, and the person responsible for usage alerts and top-ups.

Choose the subscription for the entitlements and shared capacity the people need. Choose the API model and PAYG policy for the traffic the software creates. Move to a workspace or Enterprise agreement when ownership, controls, concurrency, or contractual assurance—not just generation volume—becomes the deciding constraint.