Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.deliah.ai/llms.txt

Use this file to discover all available pages before exploring further.

Getting your Voice Clone live is straightforward. You record a few samples in different emotional styles, submit them through Deliah, and the AI does the rest. This guide walks you through each step — from preparing your space to the moment your chatters can start generating messages in your voice.
Audio quality is the single biggest factor in how realistic your Voice Clone sounds. A clean recording in a quiet room consistently outperforms a longer recording with background noise or echo.
1

Prepare your recording environment

Find the quietest space available — a closet full of clothes or a carpeted room with soft furnishings both work well. Close windows and doors, turn off fans, air conditioning units, and anything else that hums or buzzes in the background. Silence notifications on your phone before you start.For equipment, a modern smartphone held 15–20 cm from your mouth is sufficient. A dedicated USB or condenser microphone will give you noticeably better results if you have one.
2

Record your voice samples

Record at least 30 seconds of audio — but aim for 1 to 3 minutes total across all your samples. Longer recordings give the AI more material to work with and produce a more accurate, expressive clone.You need to cover three emotional variations:
  • Normal — your natural, conversational speaking voice
  • Whisper — a soft, intimate tone as if speaking quietly to someone close
  • Ecstasy — an expressive, excited, or heightened emotional delivery
Record each variation as a separate file or clip. Speak naturally within each style; scripted or stilted delivery affects the final model.
3

Submit your recordings through Deliah

Log in to the Deliah platform and navigate to your Voice Clone settings. Upload your audio files directly — WAV and MP3 formats are both accepted. You can also submit:
  • Video files (Deliah extracts the audio automatically)
  • Existing voice messages from real fan chats
  • Any other audio or video content that features your voice clearly
The more varied and high-quality material you provide, the more versatile your clone becomes.
4

Wait for Deliah to build your Voice Clone

After you submit your recordings, Deliah processes them and builds your voice synthesis model. This typically takes a short time depending on the volume of audio submitted. You’ll receive a notification when your Voice Clone is ready.You don’t need to take any action during this step — just wait for the confirmation.
5

Your chatters can now send voice messages in your voice

Once your Voice Clone is active, your chatters can generate voice messages directly from the Deliah platform. They write or select a message, the AI renders it in your voice, and it gets delivered to the fan. You don’t need to record anything new for each message.Review your first batch of generated messages to confirm the clone sounds right. If something feels off, submitting additional recordings — especially in the variation that sounds weakest — will improve the model.
Don’t start from scratch if you already have audio or video content. Existing YouTube videos, Instagram Reels, podcast episodes, TikToks, or saved voice messages from fan chats are all valid sources. Submit whatever you have — Deliah can extract your voice from most formats, and reusing existing content is the fastest way to build up recording time.