This guide walks you through creating your first voice profile and generating speech in PeachDesk Voice Studio.

​
Prerequisites

Sign in to your PeachDesk account at app.peachdeskai.com and open Voice Studio from the PeachDesk dashboard. There is nothing to install — Voice Studio runs entirely in your browser, and all models run on PeachDesk’s GPU infrastructure.

​
Step 1: Create a Voice Profile

Voice profiles are the foundation of Voice Studio. Each profile defines the voice that’s used when you generate speech. You can either clone a voice from your own audio or pick a built-in voice.
1

Open the Voices tab

Click the Voices tab to manage your voice avatars.
2

Create a new voice

Click + New Voice and fill in the details:
  • Name: A descriptive name (e.g., “John Smith”)
  • Language: Select the primary language
  • Description: Optional notes about the voice
3

Choose how to create the voice

You have two options:Option A: Clone from audio
  • Select Clone from audio
  • Upload an audio file (WAV, MP3, or M4A), or record a sample directly in your browser
  • Ideal length: 10-30 seconds of clear speech
Option B: Built-in voice
  • Select Built-in voice
  • Choose from the library of ready-to-use voices — no audio required
4

Save the voice

Click Create Voice to save it to your workspace.
For best results when cloning, use clean audio with minimal background noise and a consistent speaking tone.

​
Step 2: Generate Speech

Now let’s use your new voice to generate speech.
1

Open the Generation tab

Click the Generation tab.
2

Select a voice

Choose your newly created voice from the dropdown.
3

Select a model

Pick a generation engine from the model dropdown. Models are hosted and served on PeachDesk’s servers — there’s nothing to download or configure.
Paralinguistic tags like [laugh], [sigh], and [gasp] only work with VAANI. VAANI-ENGLISH, VAANI-MULTILINGUAL, and VAANI-INDIC have no tag tokens in their tokenizers, so they read those tags literally instead of turning them into expressive sounds.
To insert supported tags, select VAANI and type / in the text input to open the tag inserter. The recognized tags are [clear throat], [sigh], [shush], [cough], [groan], [sniff], [gasp], [chuckle], and [laugh] — anything outside that list is spoken out loud.
4

Enter text

Type or paste the text you want to generate into the input box.
5

Generate

Click Generate and wait a few seconds.
The first generation in a session may take a little longer while the model warms up. Subsequent generations are faster.
6

Play & download

  • Click Play to preview the audio
  • Click Download to save the audio file
  • The generation is also saved to your Captures

​
Step 3: Apply an Effect (Optional)

The Effects tab lets you transform a generated clip — for example, adjusting tone or adding processing. Each effect you apply is saved as a new versioned take, so your original audio is never overwritten.
1

Open the Effects tab

Click the Effects tab and select a clip from your Captures.
2

Choose an effect

Pick an effect and adjust its settings.
3

Apply

Click Apply to render a new take. You can compare takes and roll back to any previous version at any time.

​
Step 4: Build a Story (Optional)

The Stories tab lets you create multi-voice narratives with a timeline-based interface.
1

Create a new story

Open the Stories tab and click + New Story.
2

Add voice tracks

Click + Add Track to create tracks for different speakers.
3

Add audio clips

  • Drag generated audio from your Captures onto the timeline
  • Or generate new clips directly in the timeline
  • Arrange clips along the timeline
4

Edit & export

  • Trim clips by dragging their edges
  • Adjust timing and spacing
  • Click Export to render the final audio

​
What’s Next?

Voice Cloning Guide

Learn advanced techniques for high-quality voice cloning.

API Integration

Integrate Voice Studio into your own applications.

Stories

Master the multi-track timeline editor.

Talk

Have a real-time conversation in any of your voices.

​
Tips for Success

  • Use 10-30 seconds of clear, consistent speech when cloning
  • Avoid background noise and echo
  • Multiple samples from the same speaker improve quality
  • Match the speaking style you want to generate
  • Generation runs on PeachDesk’s GPU infrastructure, so speed is consistent regardless of your device
  • Reusing the same voice across generations keeps results fast and consistent
  • Keep individual text inputs reasonably sized for the quickest turnaround
  • Generation is slow or queued: Try again in a moment — heavy load is balanced across our servers
  • Poor audio quality: Try adding more (or cleaner) voice samples to the profile
  • Tags read out literally: Switch to VAANI, which is the only model that supports paralinguistic tags
  • See the full Troubleshooting Guide for more