Consented brand voices
Organizations with a named speaker's written permission to deploy a consistent synthetic voice in approved customer experiences.
Independent tool overview
xAI Custom Voices at a glance
xAI Custom Voices lets developers clone a consenting speaker's voice from a short recording, then use the resulting voice ID with Grok text-to-speech and real-time voice APIs. Console creation is available in most of the United States, while API-based creation requires an Enterprise plan.
Visit the official xAI Custom Voices site ↗
- Best for
- Developers building consented branded narration, support agents, accessibility tools, and character voices
- Availability
- United States except Illinois, according to current xAI documentation
- Reference audio
- Up to 120 seconds; xAI recommends 90–120 seconds for best results
- Consent control
- Console flow uses a live passphrase and speaker-similarity check
- Team limit
- Up to 30 custom voices by default
- API use
- Works with REST and WebSocket TTS plus real-time speech-to-speech
- Usage price
- $15 per 1M TTS characters or $0.08 per speech-to-speech minute for the current main voice model
- Last reviewed
- September 1, 2026
Overview
What xAI Custom Voices is
Custom Voices is an xAI developer feature, not a general voice-cloning tool inside the consumer Grok chat experience. A team records or uploads reference speech, creates a private custom voice, and passes its voice ID to Grok text-to-speech, streaming TTS, or real-time speech-to-speech endpoints.
xAI's launch announcement describes a two-stage ownership check in the console: the speaker reads a live passphrase, then the system compares speaker embeddings from that check and the longer recording. The company says this prevents cloning from an old clip or cloning somebody else's voice. That is a meaningful control, but it does not remove the developer's duty to obtain durable, use-specific consent and prevent deceptive output.
Current documentation recommends 90–120 seconds of clean, expressive, single-speaker audio, although clips can be shorter and may not exceed 120 seconds. Custom voices are private to the creating team, can be used anywhere built-in xAI voices work, and inherit multilingual output and streaming capabilities.
The existing Rundown description said Custom Voices were for use across Grok applications. The current first-party documentation is narrower and more useful: this is primarily an API and developer-console capability for TTS and voice agents. Treat every cloned voice as sensitive identity data, disclose synthetic speech, secure its source audio and voice ID, and never use it for impersonation or unsanctioned reuse.
Use cases
Who xAI Custom Voices is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Creator narration
Creators scaling drafts, revisions, localization, or accessible versions in their own voice while clearly labeling synthetic audio.
Voice accessibility
People preserving their own vocal identity for assistive communication, with careful long-term consent and account recovery planning.
Real-time voice applications
Engineering teams that need the same custom voice across generated narration and conversational agents.
Capabilities
Core xAI Custom Voices features
Short-reference voice cloning
Creates a reusable voice model from no more than 120 seconds of reference audio.
Two-stage console verification
The announced console flow checks a live spoken passphrase and similarity between that recording and the main reference clip.
Team-scoped voices
Custom voices appear only to the owning team and are not added to xAI's public built-in voice list.
TTS and real-time compatibility
A custom voice ID works with REST TTS, streaming TTS, and real-time speech-to-speech endpoints.
Multilingual synthesis
Custom voices inherit the multilingual output capabilities available to xAI's voice APIs.
Voice metadata
Teams can label a voice with a name, description, language, accent, age, gender, use case, and tone.
Lifecycle endpoints
Enterprise API users can list, inspect, update metadata, download reference audio, and delete a voice.
No cloning surcharge
xAI says using a custom voice carries the same TTS or Voice Agent usage charge as a built-in voice.
Process
How the xAI Custom Voices workflow works
Step 1
Document the use case
Define where the voice will appear, what it may say, who can use it, how long permission lasts, and how consent can be withdrawn.
Step 2
Obtain explicit written consent
Get permission from the actual speaker for cloning, storage, generation, distribution, languages, commercial contexts, and any downstream customer use.
Step 3
Record a clean reference
Capture 90–120 seconds of natural, expressive, single-speaker audio in a quiet, soft room with no music or private conversation.
Step 4
Complete the live verification
Use the console passphrase and speaker-similarity flow; Enterprise teams should confirm in writing how verification applies to API-created voices.
Step 5
Restrict team access
Keep API keys, console access, reference audio, and voice IDs limited to people who need them, with logging and rapid revocation.
Step 6
Test representative scripts
Review pronunciation, emotional tone, accents, multilingual output, names, numbers, and high-risk phrases before launch.
Step 7
Add disclosure and escalation
Tell listeners they are hearing an AI-generated or cloned voice and provide a clear route to a human for consequential interactions.
Step 8
Block unsafe uses
Prevent identity verification, payment authorization, political persuasion, emergency instructions, deceptive endorsements, and unsupervised high-stakes advice.
Step 9
Monitor and revoke
Audit generated scripts and distribution, respond to misuse, honor consent withdrawal, delete the voice and source audio, and remove deployed assets where possible.
Cost
xAI Custom Voices pricing and free plan
Creating and using a custom voice does not add a separate cloning surcharge, but generated audio is billed at the applicable xAI Voice API rate. Console users can create up to 30 custom voices; direct API creation is restricted to Enterprise teams. Prices below were checked in xAI's current API documentation.
Console voice creation
No extra creation charge
Create up to 30 custom voices in the xAI console, subject to availability and account access.
- Available only in the United States except Illinois
- Custom voices are scoped to the creating team
- Generated audio is billed separately at standard Voice API rates
Text to Speech
$15 per 1M characters
Usage-based synthesis through xAI's REST or streaming TTS endpoints.
- Same published rate for built-in and custom voices
- Actual cost depends on script length and repeated generation
- Confirm current model availability in the team's console
Speech to Speech
$0.08/minute audio
Current pricing for grok-voice-think-fast-2.0, equivalent to $4.80 per audio hour.
- Text-input charges may also apply
- Real-time application infrastructure and telephony costs are separate
- A deprecated older model is listed at a lower rate and should not anchor a new deployment
Enterprise API creation
Contact sales
Direct creation through the custom-voices API requires an Enterprise-enabled team.
- Only the console supports ordinary self-serve creation
- Request higher limits when more than 30 team voices are required
- Obtain written answers about consent, retention, support, and regional access before production
Pricing checked . Check current pricing at the source ↗
Assessment
xAI Custom Voices strengths and limitations
Where it stands out
- One custom voice ID works across TTS and real-time voice endpoints
- The console's live passphrase and similarity checks are stronger than accepting an arbitrary old recording
- Team scoping prevents a custom voice from appearing in the public built-in voice catalog
- A 90–120 second reference can capture useful vocal identity without a long studio session
- Supports both batch-like narration and interactive voice-agent experiences
- No separate custom-voice surcharge simplifies cost comparison with built-in voices
- Lifecycle APIs include reference-audio download and full deletion
- The feature can support accessibility and personal voice preservation when the speaker controls consent
What to consider
- Custom Voices is currently limited to the United States and unavailable in Illinois
- The launch post describes verification for every voice, while the current Enterprise create-API documentation lists only a reference file as required and shows no passphrase field; Enterprise buyers should resolve that discrepancy before use
- Technical ownership checks do not replace explicit, written, use-specific, revocable consent
- Never clone a person who is absent, deceased, incapacitated, employed under conflicting terms, or represented by an estate or union without the required authority
- A voice clone can enable fraud, scams, fake endorsements, political manipulation, harassment, defamation, and deceptive impersonation
- xAI's Acceptable Use Policy prohibits deceptive impersonation, privacy and publicity-right violations, fraud, fabricated identities or evidence, and misleading people about AI use
- Synthetic output should be disclosed wherever a reasonable listener could believe the real speaker recorded or approved the exact words
- Do not use a cloned voice as biometric authentication or as proof that a person approved a payment, contract, medical instruction, or other consequential action
- In the United States, AI-generated voice calls fall under the FCC's artificial or prerecorded voice rules; consent, identification, and other TCPA requirements can apply
- Voice, likeness, biometric, publicity, labor, consumer-protection, election, and recording laws vary by jurisdiction and use case
- Deleting the voice from xAI does not automatically recall audio files already generated, downloaded, cached, syndicated, or published elsewhere
- The owning team can download the original reference audio, making team permissions and API-key security especially sensitive
- Compromised credentials could expose a high-value voice identity or allow unauthorized generation
- Reference recordings can reproduce room echo, background noise, reading style, pronunciation issues, and unwanted vocal mannerisms
- Short clips under 30 seconds may lack detail; even the recommended 90–120 seconds cannot capture every emotion, language, or speaking context
- Cross-language output can mispronounce names, change accent characteristics, or imply that the speaker personally speaks a language they do not
- Cloned speech can sound natural while delivering hallucinated, outdated, unsafe, or unauthorized text from the application behind it
- Customer-support agents need human escalation and must not pretend to be the real speaker
- xAI says API inputs and outputs are retained for 30 days by default for abuse auditing; Zero Data Retention has separate eligibility and setup
- Voice API, telephony, storage, monitoring, and application-model charges can all contribute to production cost
- The default 30-voice team limit and U.S.-only availability constrain global or large-catalog deployments
- Enterprise API access, higher limits, regional compliance, service commitments, and custom safeguards require sales coordination
Compare
xAI Custom Voices alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Marketing
ElevenLabs
Choose ElevenLabs for a broader mature voice platform with multiple cloning tiers, creator tooling, dubbing, and a large ecosystem.
Explore ElevenLabs →Consumer
Qwen3-TTS CustomVoice 1.7B
Choose Qwen3-TTS CustomVoice when model access and self-managed experimentation matter more than a hosted, verified cloning workflow.
Explore Qwen3-TTS CustomVoice 1.7B →Content Creator
ElevenLabs Voice Design V3
Choose ElevenLabs Voice Design V3 when a new synthetic character voice is safer and more appropriate than copying a real person's identity.
Explore ElevenLabs Voice Design V3 →Questions
xAI Custom Voices FAQs
What is xAI Custom Voices?
It is an xAI developer feature that creates a private reusable voice from a short reference recording for use in Grok text-to-speech and real-time voice APIs.
Is Custom Voices inside the regular Grok chat app?
The current official documentation presents it as an xAI console and API capability for developers, not a general consumer cloning button across Grok chat.
Where is xAI voice cloning available?
As of September 1, 2026, xAI says Custom Voices is available only in the United States, excluding Illinois.
How much reference audio do I need?
The maximum is 120 seconds. xAI accepts shorter clips but recommends 90–120 seconds; clips under 30 seconds may capture less detail.
Can I clone someone else's voice?
No. xAI says the console requires a live passphrase and speaker-similarity check and that users cannot clone another person's voice. You also need explicit legal permission for the full planned use.
How much does xAI Custom Voices cost?
xAI says there is no extra custom-voice surcharge. Current usage is $15 per 1 million TTS characters or $0.08 per speech-to-speech audio minute for grok-voice-think-fast-2.0, with possible additional charges.
How many voices can a team create?
The default limit is 30 custom voices per team. xAI asks larger deployments to request a higher limit.
Can I create a custom voice through the API?
Only an Enterprise-enabled team can call the creation endpoint directly. Other eligible users create voices in the xAI console and then copy the voice ID.
Are custom voices public?
No. xAI says they are scoped to the creating team and never appear in the built-in voice list available to other users.
Can a custom voice speak other languages?
xAI says custom voices inherit multilingual TTS capabilities. Test every target language with the speaker because pronunciation, accent, meaning, and implied endorsement can change.
Do I need to disclose that the audio is AI-generated?
Yes whenever a listener could reasonably think the person recorded or approved the exact message. Disclosure is also a core fraud-prevention control, and specific laws or platform rules may require it.
Can I use a cloned voice for phone calls?
Technically, it can power a voice agent, but U.S. AI-generated calls are subject to FCC artificial or prerecorded voice rules. Obtain appropriate consent, provide required identification, disclose automation, and get legal review for telemarketing or scaled calling.
What happens when I delete a custom voice?
xAI says deletion removes the voice and its underlying reference audio and causes future calls with that voice ID to fail. It does not erase audio already exported or published outside xAI.
Does xAI train on voice API data?
xAI's API security FAQ says it does not train on API inputs or outputs without explicit permission. API requests and responses are stored for 30 days by default for abuse auditing, with separate Zero Data Retention options for eligible teams.
Bottom line
Our xAI Custom Voices verdict
xAI Custom Voices is a capable developer primitive with unusually clear console ownership checks, simple TTS pricing, and one voice ID across narration and live agents. The page's original description understated the product's API focus. Its safety depends on the deployment: obtain written and revocable consent, resolve the Enterprise API verification discrepancy, disclose synthetic speech, secure the source audio and credentials, and block every use where a listener could be deceived or harmed.
Visit xAI Custom Voices website ↗Similar Miscellaneous Tools
View all Miscellaneous AI tools →
Deep Max
Deep Max - Exa's new SOTA agentic search tool
Realtime TTS-2
Realtime TTS-2 - Inworld AI's new voice model that hears conversation audio to match user tone and emotion
HY-World 2.0
HY-World 2.0 - Tencent's open-source world model that turns text, images, or video into interactive 3D scenes
GPT-Realtime-2
GPT-Realtime-2 - OpenAI's new voice model that can think, call tools, and recover from interruptions in live calls
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.