ElevenLabs Development Services

AlphaCorp AI deploys five ElevenLabs models in production work, chosen per use case. The earlier Turbo v2.5 and v2 models have been superseded by the Flash line, so any code still targeting Turbo is on our migration list.

Eleven v3 handles multi-character audio and audiobooks in batch, across 70-plus languages, at 5,000 characters per request. Multilingual v2 carries emotionally consistent narration in 29 languages at 10,000 characters. Flash v2.5 is the real-time and bulk model, targeting about 75ms across 32 languages at 40,000 characters.

Scribe v2 transcribes with word-level timestamps and diarization across 90-plus languages in batch, and Scribe v2 Realtime does the same under 150ms for live agents and meetings. A v3 Conversational variant targets about 280ms for real-time agents that need v3's expressiveness, and Eleven Music v2 is available for commercial scored audio, which we scope separately.

There are two costs: AlphaCorp AI's build fee, which scope sets and a working session prices, and the ElevenLabs subscription itself, which runs from $0 on the Free tier to $990 a month on Business before custom Enterprise pricing, per ElevenLabs' pricing page. Annual billing gives about two months free.

Credits burn at different rates by product, which is where budgets go wrong: text-to-speech runs roughly 0.5 to 1 credit per character, with Multilingual v2 costing more per character than Flash; speech-to-text runs about 330 credits a minute; dubbing runs 2,000 to 10,000 credits a minute depending on tier; and music generation runs about 900 credits a minute.

What moves our fee is the number of channels, phone, app, and web, the depth of integration with your CRM and knowledge base, and whether Enterprise controls such as regional data residency are in scope. Regulated deployments cost more to build and less to fix later.

We build on the platform's Enterprise controls and add the disclosure and consent logic that US and EU law now require.

Platform certifications. ElevenLabs documents SOC 2 Type II, ISO 27001, ISO 27701, ISO 42001, and PCI DSS Level 1, with HIPAA Business Associate Agreements for qualifying healthcare customers and GDPR attestations. Those are ElevenLabs' certifications, and we design so your deployment stays inside their boundary.

Where your audio goes. Enterprise customers can select US, EU, or India regional environments, and EU residency keeps data on isolated EU infrastructure. Every other plan processes data on default US infrastructure regardless of where the user sits. For a European buyer, that fact alone decides the plan tier.

Disclosure on calls. The FCC ruled in February 2024 that AI-generated and cloned voices count as artificial under the Telephone Consumer Protection Act, so outbound voice agents carry the same consent, disclosure, and opt-out obligations as any robocall. We build those into the call flow.

Marking and consent. Under the EU AI Act's Article 50 transparency obligations, enforceable since August 2026, synthetic audio must carry machine-readable marking and deployers of audio deepfakes must disclose them. ElevenLabs embeds imperceptible watermarks and participates in C2PA; we make sure your product keeps that provenance intact and adds the audible disclosure where required. Every cloned voice needs documented, specific consent before it enters a build.

One caution. A September 2025 benchmark study on adversarial attacks against audio deepfake detectors found that detectors trained on the ASVspoof 2019 corpus generalize poorly to current synthetic speech. Detection is a weak control. Disclosure and provenance are the ones we rely on.