Connect ElevenLabs to HeyGen · Secure 2026 Voice Guide

Quick Summary

• Create a dedicated, restricted ElevenLabs API key, then import it from HeyGen's AI Studio through Voice, New Voice, Integrate 3rd Party Voice and Import Voice. Choose ElevenLabs, paste the key and confirm.
• HeyGen currently says the integration may need access to Text to Speech, Voices, Voice Generation, Models, History and User. Start with only the documented permissions and a credit limit instead of turning off every restriction.
• An API key is a secret with access to your ElevenLabs account resources and quota. Never place it in a prompt, screenshot, public repository, browser-side code or shared document. If it is exposed, rotate it immediately.
• After connecting, run a short test with one approved voice and a diagnostic script. Check voice identity, pronunciation, pacing, lip sync and credit usage before moving a client or publishing workflow onto the integration.
• If you do not want HeyGen to retain a working ElevenLabs credential, generate and approve the narration in ElevenLabs, then upload the audio file to HeyGen instead. The tradeoff is more manual handling in exchange for a smaller credential boundary.
• The video was published on July 15, 2026. Product interfaces can change, so the steps below are checked against current official documentation available through October 21, 2026.

What the ElevenLabs to HeyGen connection actually does

The integration lets HeyGen request speech from eligible voices in your ElevenLabs account. You write or paste the script in HeyGen, select the imported voice and generate an avatar scene without exporting a separate narration file for every revision.

The API key is the bridge. It authenticates HeyGen's requests and charges the resulting usage against the ElevenLabs workspace tied to that key. HeyGen's own credits and limits still apply to avatar video generation, so this is a two-account production chain rather than one combined subscription.

Credit: the embedded walkthrough belongs to Learn With Trevor. The exact setup below combines the video's accessible explanation with HeyGen's current third-party voice guide and ElevenLabs' current API-key security documentation.

Before you connect the accounts

Open ElevenLabs and HeyGen in separate tabs and confirm that you are in the correct workspaces. If you work with clients, stop here and decide who owns the voice, the ElevenLabs workspace, the HeyGen workspace and the generated files. A personal key should not quietly become a permanent credential for someone else's production system.

Confirm that the voice is available to the account and allowed for the intended use. ElevenLabs requires users creating an Instant Voice Clone to confirm that they have the right and consent to clone the voice. A technical connection does not grant new rights to a voice or likeness.

Choose one low-risk voice and a short test script before connecting. Do not start with a client's launch video, a long paid render or sensitive material. The first goal is to prove authentication, voice visibility, speech generation and cost tracking.

Step 1: create a dedicated ElevenLabs API key

In ElevenLabs, sign in, open Developers in the left sidebar and select API Keys. Create a new key instead of reusing a general-purpose credential. Give it an unambiguous name such as HeyGen production or HeyGen test so later rotation and usage review are straightforward.

ElevenLabs shows the full secret only when the key is created. Copy it directly into the official HeyGen connection field when you reach that step. If you lose the secret, create a replacement rather than storing it in a note, prompt, screenshot or source file for convenience.

Keep Restrict Key enabled. HeyGen's current troubleshooting guide lists Text to Speech, Voices, Voice Generation, Models, History and User as the permissions that may be necessary. Start with those documented capabilities. ElevenLabs also supports a per-key credit limit and, for user keys, an expiry. These controls reduce the impact of misuse or a forgotten integration.

ElevenLabs documents service-account keys for multi-seat workspaces and recommends them for production automation because they are not tied to one employee. A user key is acceptable for an individual setup, but a team should avoid making a long-lived production connection depend on one person's membership.

Step 2: import the ElevenLabs voice in HeyGen

The current official HeyGen path begins on the home page. Hover over Create, choose Create a Video, then Create in AI Studio. Select the avatar for the scene, hover over the person-shaped Scene icon and open the Voice dropdown.

In the voice settings panel, choose Voice, then New Voice. Select Integrate 3rd Party Voice, choose Import Voice and select ElevenLabs as the provider. Paste the API key into the designated field, check that no spaces were added before or after it, then choose Confirm.

Eligible ElevenLabs voices should now appear in HeyGen's voice list. HeyGen also supports importing from Avatar Settings and its Proofreading Tool, but AI Studio is the clearest first test because it keeps the avatar, script and voice in one scene.

Do not paste the secret into HeyGen's script box or any AI chat. The only appropriate destination in this workflow is the platform's dedicated API-key field on the official HeyGen domain.

Step 3: verify the voice before real production

Create a short 10 to 15 second scene using the imported voice. Include a normal sentence, your brand or creator name, a number, an acronym and a question. This reveals pronunciation and pacing problems without consuming the credits required for a full video.

Confirm five things: the expected voice appears, HeyGen can generate speech, the avatar lip sync follows the audio, ElevenLabs records the expected usage and neither account reports a quota or permission error. Save the test result with the key name and date, but never save the key value with it.

A successful authentication test is not a content approval. Review the exact generated voice, the avatar performance, captions and factual claims before publication. Use a human voice or avatar owner to approve any public-facing clone when identity fidelity matters.

Troubleshooting the most common connection failures

Invalid API key: copy the secret again from the creation step, remove accidental spaces and confirm that it has not expired, been deleted or been disabled. If the full value is no longer visible, make a new restricted key. Do not repeatedly weaken an existing key without understanding which permission is missing.

Authentication works but no voice appears: verify that the voice belongs to the connected ElevenLabs account or workspace, that you have permission to use it through the API and that Voices, User and Models access are enabled. Refresh HeyGen after changing key permissions.

The voice appears but generation fails: check ElevenLabs credits, the key's credit ceiling, HeyGen credits and any account or model restrictions. HeyGen specifically notes that integrated voices stop working when the connected third-party account runs out of credits.

The wrong voice is selected: reopen the voice list, compare the voice name and run the diagnostic script again. Do not rely on a similar preview alone. For client work, keep a small record of the approved voice name, owner, language, model and test sentence.

A key works and then suddenly stops: check expiry, deletion, workspace membership and secret-scanning alerts. ElevenLabs says keys found in a public GitHub repository can be disabled automatically. Treat that event as a security incident, not as a reason to paste the same secret somewhere else.

Treat the API key like a password with spending power

An ElevenLabs API key can authenticate requests and consume workspace quota. The official documentation says not to share it, expose it in client-side code or commit it to a repository. The safest practical setup is one dedicated key per integration, the minimum documented scopes, a credit limit and an owner who knows how to revoke it.

Do not put the key in a Claude, ChatGPT or support prompt, even when asking for troubleshooting help. Share only the error message, key name, last four characters, selected scopes and account state. A support screenshot should crop or mask any secret field.

For a shared production workflow, use a service account when the ElevenLabs plan supports it. Separate testing and production. Keep client workspaces isolated, log who approved the voice and review usage regularly. IP restrictions can strengthen some direct API deployments, but a hosted third-party integration may use changing infrastructure, so do not add an IP allowlist unless the provider publishes compatible addresses.

How to rotate or revoke the connection

Rotate a healthy key by creating a replacement with the same necessary permissions, updating HeyGen to use the new key, testing one scene and only then deleting the old key. This sequence avoids unnecessary downtime while keeping the old credential alive for the shortest practical period.

If a key may be exposed, delete or disable it immediately. Do not wait for the replacement test. Create a new restricted key, reconnect HeyGen, review recent usage and document what was exposed. ElevenLabs says deleting a key is irreversible, which is appropriate for a compromised secret.

When a project ends, remove the imported voice or integration from HeyGen where available and revoke the dedicated key in ElevenLabs. A disconnected-looking interface is not proof that the secret can no longer authenticate. Revocation at the credential issuer is the decisive control.

Alternative: upload approved ElevenLabs audio instead

A direct connection is convenient for frequent revisions. An audio-file handoff is easier to audit. Generate the final narration in ElevenLabs, approve its pronunciation and delivery, download it, then upload that exact file to the HeyGen scene as custom audio.

This method avoids giving HeyGen a persistent ElevenLabs key and makes the approved audio the timing master for the avatar. It is useful when a security policy forbids third-party credentials, when different teams own the voice and video stages, or when a client wants to approve narration before avatar generation.

The tradeoff is manual work. Every script change requires a new audio export and upload. Choose the direct integration for a trusted, frequent production loop. Choose the audio handoff for tighter review boundaries, occasional videos or sensitive accounts.

Production checklist

Before connection: confirm account ownership, voice rights, workspace, plan access and the person responsible for costs. Create a dedicated key with the documented scopes, a credit limit and an expiry when appropriate.

Before generation: confirm the selected voice, run the diagnostic script, verify both credit balances and approve the script. Keep sensitive information out of the text unless both platforms and the client's data policy permit it.

Before publication: review pronunciation, lip sync, captions, claims, consent and disclosure. Before offboarding: export required records, remove the integration from the project, revoke the ElevenLabs key and confirm that no automation still depends on it.