Skip to main content
You can add a voice of your own, for example a presenter’s or your brand’s voice, from a short recording. The voice is cloned from 3 to 20 seconds of clear speech. Once created, it is a project voice: it appears under Project voices in the Voices library, and “Everyone in this project can use it.”

Before you start

  • You need to be an owner or admin, or have the voices permissions. See Roles and permissions.
  • Only clone a voice you have the right to use.
  • Have a recording of one person speaking clearly, at least 3 seconds long, with little background noise.

Add the voice

1

Open Add new voice

In Voices library, click Create a voice. You can also start from Clone a voice on Home, or from Clone voice in the voice picker of Voice overs. The same form opens.
2

Fill in the voice details

In Voice Details, set:
  • Cover image: optional. JPG, JPEG or PNG, at most 5 MB.
  • Name: required. Letters and numbers in English or Arabic, spaces, @ and -, up to 190 characters.
  • Language: Arabic (the default) or English.
  • Dialect: required for Arabic. For English it says “Not required for English”.
  • Tags: “Select one gender, one style, and at least one use case”. Pick Male or Female, one Style (such as Narrator or Conversational), and one or more Use cases. These are what the library’s filters use to find the voice.
Add new voice, step 1 Voice Details: name filled in, Arabic with the Saudi Arabia dialect, Female, Conversational, two use cases, the audio ready and the transcription confirmedAdd new voice, step 1 Voice Details: name filled in, Arabic with the Saudi Arabia dialect, Female, Conversational, two use cases, the audio ready and the transcription confirmed
3

Add the audio

Under Input audio, choose Upload or Record.
  • Upload: click Add or drop your audio file, or drop a file on it. “MP3, WAV, WebM, OGG, AAC, M4A, FLAC • Max 35MB”
  • Record: click Start recording, speak, then Stop recording. “Record 3 seconds or more; over 20, you pick the part.”
Input audio set to Record, with the Record your voice area and its hint about 3 to 20 secondsInput audio set to Record, with the Record your voice area and its hint about 3 to 20 seconds
4

Pick the part to clone from, if asked

If the audio is longer than 20 seconds, Pick the part to clone from opens. “A voice is cloned from 3 to 20 seconds: drag the handles or the selection.” The first 20 seconds are selected to start with. Click Play the part to hear it, then Use this part.
Pick the part to clone from: the waveform of a 30-second file with the first 20 seconds selected between two handles, Choose another file and Use this partPick the part to clone from: the waveform of a 30-second file with the first 20 seconds selected between two handles, Choose another file and Use this part
5

Confirm the transcription

The Transcription fills in on its own from the audio (“Extracting transcription…”). Read it and correct any word that differs from what was said, then tick Confirm transcription. You can’t continue until it is confirmed.
6

Preview the voice (optional)

Click Next. In Generate Custom Template, the Text box has an example sentence in the voice’s dialect. Keep it, click Example for another, or write your own (at least 5 words). Click Generate to hear how the new voice sounds.
Step 2 Generate Custom Template: an Arabic example sentence in the Text box, Example and Generate buttons, and Previous and Create at the bottomStep 2 Generate Custom Template: an Arabic example sentence in the Text box, Example and Generate buttons, and Previous and Create at the bottom
7

Create it

Click Create. “Voice created successfully”. The voice is under Project voices in the Voices library.

Use the trimmer

The trimmer shows the waveform of your file or recording with the selected part between two handles.
  • Drag a handle to move the start or the end, or drag the selection to move it whole. The part stays between 3 and 20 seconds.
  • With a handle focused, the arrow keys move it by a tenth of a second, and Shift with an arrow by one second.
  • Choose another file (or Record again for a recording) goes back to pick other audio. Escape returns to the form.
  • The chosen part is cut in your browser and uploaded in place of the whole file.
Pick a part where only the speaker talks, with no music, laughter or long pauses.

Audio limits

Arabic and English

  • An Arabic voice needs a dialect, and it speaks that dialect. Choose the one the speaker in the recording uses.
  • An English voice has no dialect.
  • The transcription is read in the language you chose. If you change the language after adding the audio, the transcription is cleared and read again, and you confirm it again.

Good to know

  • You can’t save a voice that is identical to one you already added in this project (same name, dialect and gender): “Identical TTS voice already exists for this account! Try changing some values.”
  • A voice can’t be edited after it is created. To change it, add it again and delete the old one.
  • Closing the dialog with details filled in asks before discarding them. Switching project closes it.
  • The new voice also appears in Hamsa Agents for the same project. See Add a voice in Hamsa Agents.

Voices library

Find, compare and delete project voices.

Generate speech

Use your new voice in Text to speech.

Dictionaries

Fix how the voice says specific words.