Skip to main content
Skip to workspace

Reference Voice Text to Speech

Loading generation options…

Prepare the sample, consent and words separately

Illustration of preparing a clear voice recording
Recording preparation illustration.

Input example

One speaker introduces a project

Example script

Welcome to our project. In this short introduction, I will explain what we are making, who it is for and where you can follow the next update. Thank you for taking a moment to listen.

About this example
What you provide
  • A clear WAV sample of the adult speaker’s own voice, within the displayed limits.
  • A separate WAV recording of that speaker reading the exact consent statement displayed in the form.
  • The words you want this generation to speak.
What to expect

Speech guided by this request’s reference; it does not create a reusable saved voice.

What to check
  • Keep the sample and consent files in their respective slots.
  • Inspect pronunciation and likeness without assuming an exact voice copy.

Input example

Two speakers keep separate reference files

Speaker turns
  1. Speaker one

    I have the first version ready. Would you like to hear the opening?

  2. Speaker two

    Yes. Let us listen to the whole passage before deciding what to change.

Illustrative role portraits.

About this example
What you provide
  • Speaker one: their own sample and their separate consent recording.
  • Speaker two: their own sample and their separate consent recording, when the selected model supports a second speaker.
  • Dialogue lines assigned to the correct speaker.
What to expect

A supported one- or two-speaker reference-based recording.

What to check
  • Do not swap sample and consent recordings or reuse another speaker’s consent.
About these examples

This is a reference-voice input example, not a saved voice or a generated audio result. Use the same adult speaker’s own voice for the sample and separate consent recording required by the form.

A CLOSER LOOK

Reference Voice Text to Speech

Generate speech with a voice reference or a preset voice. Reference mode uses an adult’s voice sample and separate consent recording alongside your script.

Reuse goal
This reference applies to the current task and does not create a saved speech voice.
Sample quality
One clear speaker, consistent recording conditions and limited background sound.

Reference Recording, Consent and Script Are Separate

ItemIts roleCheck before submission
Voice sampleGuides the voice for this generationThe accepted sample from the intended adult speaker
Consent recordingSupplies the separately required statementThe same speaker reads the exact statement shown in the form
ScriptContains the words to generateCorrect role assignments and the text you want heard
A second reference-based speakerAdds another supported roleIts own required sample/consent pair, rather than reusing the first speaker’s files
An adult records a clear solo spoken voice sample at a single microphone in a quiet booth.
AI-created editorial illustration.
01

Reference Voice Text to Speech for This Recording

Reference Voice Text to Speech uses an accepted preset or an authorized reference recording alongside the words to generate. In reference mode, each speaker needs the required voice sample and the separate consent recording. This is a way to guide a particular spoken-audio task; it does not automatically create a reusable voice in your saved library.

Prepare the adult speaker’s own clear voice and follow the exact consent text shown in the form. The current reference workflow asks for a 10–30 second WAV voice sample plus a separate WAV consent recording from the same speaker. Review the displayed limits before recording because the two files have different purposes.

A recording kit groups two separate recorders with a script folder to prepare distinct reference and consent recordings.
AI-created editorial illustration.
02

Keep Voice Samples, Consent and Script Paired Correctly

Use a quiet recording with one speaker and avoid background music or overlapping conversation. Keep the reference and consent files paired with the correct role. Consent material does not replace the voice sample, and a familiar-sounding preset is not evidence that you have permission to use someone else’s private recording.

Write the actual script and choose one or two speakers through the available controls. For a second reference-based role, supply its own required pair of recordings. Check all role assignments before generating so a change to one speaker does not leave you evaluating the wrong reference against the resulting voice.

A listener compares a source voice recording with a new narration take using headphones and two physical audio players.
AI-created editorial illustration.
03

Review Reference Voice Similarity and Spoken Delivery

Listen to the output for wording, pacing, language and the vocal qualities that matter to the task. A reference guides generation without guaranteeing an exact identity match or the same similarity for every script. Difficult pronunciation and large differences in recording conditions can change how convincing a take sounds.

Download the delivered audio after checking the full recording. Reference-based speech and a trained singing voice are separate workflows, so choose the one that matches your recording.

How to Use Reference Voice Text to Speech

  1. 01

    Choose a reference recording or preset voice

    For reference mode, provide a 10–30 second WAV voice sample and a separate WAV recording of the displayed consent statement. Both recordings must use the same adult’s own voice; follow the file limits shown in the form.

  2. 02

    Write the speech for one or two speakers

    Enter the words to generate, or add a second speaker and write the dialogue turns. Each speaker can use a supported preset or the required pair of reference recordings.

  3. 03

    Review the reference-based speech

    Listen for pronunciation, pacing and voice likeness. References apply to this generation and do not create a saved voice.

Reference Voice Text to Speech Use Cases

  • Develop an authorized reference-guided spoken passage.
  • Compare a preset with a reference for a specific script.
  • Create a short one- or two-speaker recording.

Reference Voice Text to Speech Questions

Does reference voice create a saved cloned voice?

No saved voice is created by this workflow. The reference applies to the current generation.

Why are both a sample and a consent recording required?

The reference workflow uses separate recordings for vocal guidance and the displayed consent statement. Both must come from the same adult speaker’s own voice and meet the form’s requirements. Uploading one file into both roles is not a substitute for following those instructions.

Can I generate a conversation using reference voices?

The available workflow supports one or two speakers. Each reference-based speaker needs its own accepted sample and consent pair, or a supported preset where that option is available. Review the role assignments and resulting speaker turns before use.

How much does this task cost?

Review the credit cost displayed for the selected model and output settings before submitting. Available controls, limits and prices depend on that selection. You can prepare the inputs before signing in; submission requires the account and credit checks shown by the workspace.

What should I check before sharing or keeping a result?

Use material you own or have permission to transform and review the applicable service and model terms. Inspect the entire returned file before publishing. Download the results you want to keep; do not treat temporary access or a saved task record as a promise of permanent file storage.

Open Reference Voice Text to Speech