跳转到主要内容
You can create a custom voice model without writing any code by using the Qhaigc online lab at lab.qhaigc.net. Once your voice is ready, you use it in the same way as any other voice character — pass its name as the voice parameter in POST /v1/audio/speech.
Prefer automation or need to create voice models programmatically at scale? Use the POST /v1/voice-clone/create-model API endpoint instead.

Steps

1

Log in to the online lab

Go to lab.qhaigc.net and sign in with your Qhaigc account credentials.
2

Open the Voice Clone section

In the left sidebar, navigate to Voice Clone and then click Create Voice Model.
3

Upload your reference audio

Upload 2–3 audio files that clearly capture the voice you want to clone. The system accepts common formats such as .wav, .mp3, and .m4a.
4

Enter the voice details

Fill in a name for the voice character. This is the value you will use as voice in TTS API calls. Optionally add a description to help you identify the voice later.
5

Set visibility

Choose whether to make the voice public (available to all Qhaigc users) or private (only your account). Private voices require a contact email.
6

Start training

Click Start Training. Processing usually completes within a few minutes. You will receive a notification when the voice is ready.
7

Use the voice in TTS requests

Once training is complete, call GET /v1/audio/voices to confirm the voice appears in your list, then pass its name as the voice parameter in POST /v1/audio/speech.

Tips for better results

Use recordings that are 10–60 seconds long. Shorter samples may produce less accurate clones; longer samples do not significantly improve quality beyond 60 seconds.
  • Record in a quiet environment with no background music, echo, or noise.
  • Use a single speaker throughout all uploaded files.
  • Speak naturally and clearly — avoid whispering, shouting, or extreme emotional variation.
  • Higher-quality audio (16 kHz or 24 kHz mono .wav) gives the best results.

Web console vs. API

Web consoleAPI endpoint
Requires codingNoYes
Batch creationNoYes
Best forOne-off clones, experimentationAutomation, production workflows