How to Clone Your Voice with Codex

Codex can coordinate a complete voice-cloning workflow from your project or terminal. Connect the CreativeClaw MCP server, provide a local recording, and ask Codex to create a reusable ElevenLabs or Cartesia clone. The saved voice can then narrate demos, release videos, tutorials, or any other approved script.

Step 1: connect CreativeClaw to Codex

Add CreativeClaw as a remote MCP server in Codex and complete the CreativeClaw sign-in flow. See the connection guide for the current MCP URL and setup instructions.

Once connected, ask Codex to list the CreativeClaw tools or Characters to confirm that the connection is active.

Step 2: prepare a strong recording

Record 1 to 2 minutes of clean solo speech in your primary language. Include a few natural shifts in delivery:

• relaxed conversation
• upbeat enthusiasm
• a serious explanation
• a warm welcome
• a clear call to action

Record in a quiet space. Keep the microphone position stable and avoid background music, clipping, echo, or other speakers. Save the file in a common audio format and note its local path.

Step 3: ask Codex to upload and clone the voice

Give Codex the exact file path and a scoped instruction:

Use CreativeClaw to upload /absolute/path/to/my-voice-sample.wav as a private
voice-cloning sample. Create a reusable voice-only Character named "My Voice"
and clone it with ElevenLabs. This is my own voice, and I explicitly consent
to its processing and cloning for use in my CreativeClaw workspace.

Replace the sample path with the real local path. If you want Cartesia, name Cartesia instead. Codex should use the private voice upload flow, obtain a private asset ID, and then attach the clone to a reusable Character.

Do not ask Codex to publish the source recording or turn it into a public URL. A voice sample should stay on the private cloning path.

Step 4: choose the provider

Provider	Best starting point for
ElevenLabs	Expressive narration, polished voiceovers, and broad language support
Cartesia	Fast Character speech with direct emotion, speed, and volume controls

CreativeClaw supports both providers. The provider is chosen explicitly, so a failed request should not silently switch to the other service.

Step 5: generate a speech

After the clone is ready, ask Codex:

Use the CreativeClaw Character named "My Voice" to generate speech saying:
"The new build is ready. In this release, we improved rendering speed,
simplified setup, and fixed the most common export issue."
Use a calm, confident delivery and return the generated audio.

Codex can also read a script from a workspace file:

Read /absolute/path/to/release-script.md, make it sound natural when spoken,
then generate it with my saved CreativeClaw voice. Do not change technical
names or version numbers.

Reuse the clone safely

The voice is saved on a Character, so future tasks can use the Character ID without re-uploading the sample. Ask Codex to list your Characters if the name is ambiguous.

Voice cloning requires explicit, informed permission. File access does not prove consent, and an audio recording should never be treated as automatic authorization. Review the voice-cloning terms and privacy details before cloning another speaker.

Frequently asked questions

How do I clone my voice with Codex?

Connect CreativeClaw as a remote MCP server, give Codex the path to a clean 1 to 2 minute recording, request a private voice-sample upload, choose ElevenLabs or Cartesia, explicitly confirm speaker permission, and save the result on a reusable Character.

Can Codex create a voice clone?

Yes, when CreativeClaw is connected as a remote MCP server. Codex can upload the recording through CreativeClaw’s private voice-sample flow, request an ElevenLabs or Cartesia clone, and generate speech with the saved Character.

What audio should I give Codex?

Use a clean 1 to 2 minute recording of one speaker in the speaker’s primary language. Include several natural emotional tones and avoid music, echo, overlapping voices, and aggressive audio processing.

Should I use ElevenLabs or Cartesia?

ElevenLabs is a strong choice for polished and expressive narration. Cartesia is useful for fast Character speech with direct delivery controls. CreativeClaw supports both, and Codex can help you choose.

Is the original recording public?

No. When it is uploaded with the voice-cloning purpose, CreativeClaw stores it as a private sample and returns a private asset ID rather than a public media URL.

Can Codex reuse the voice in another project?

Yes. The clone is attached to a CreativeClaw Character in your workspace. Ask Codex to list your Characters and use the saved Character ID when generating speech.

How much does Codex voice cloning cost?

CreativeClaw currently does not charge credits for creating a reusable clone, but the workspace must have made a credit purchase. Generating speech with the saved Character uses credits based on the selected speech model.

Can a Codex voice clone speak multiple languages?

Yes, when the chosen ElevenLabs or Cartesia speech model supports those languages. Record a strong sample in the speaker’s primary language and test pronunciation with a short target-language script first.

How do I remove the voice clone from CreativeClaw?

Ask Codex to delete the CreativeClaw Character that owns the clone. This removes its associated provider voice copies and private source samples.