Talking head

Before you start

• An authorized portrait or compatible source video.
• Speech audio or the supported voice input for the selected model.

How to use this workflow

1. Step 1

   Open Talking Head and choose the source type supported by the model.

2. Step 2

   Add the portrait/video and speech input, then review the available settings.

3. Step 3

   Check lip synchronization, facial movement, audio clarity, and the complete duration.

What to check

Only use a person’s likeness and voice with appropriate authorization. Input support differs by talking-head model.

Prepare the inputs

• Use a portrait or source video whose subject is clear and whose input type is supported by the model. Confirm authorization for both the likeness and voice.
• Prepare clean speech with the intended pronunciation and pauses. Listen for clipping, background noise and a cut-off ending before using it as input.

Choose the right controls

In Talking Head, check the chosen model’s portrait/video, audio and duration requirements. Prepare narration first when you need a controlled script. Do not treat native audio in a general video model as the same input workflow.

Troubleshoot the result

1. Lip sync or facial movement looks wrong

   Check whether the supplied speech is clear and the face is unobstructed. Review the entire result, including closed-mouth pauses, teeth and head turns. Simplify the source framing before trying again.

2. Speech is fluent but inaccurate

   Correct the source audio or narration before another animation attempt. Have a speaker of the target language check names, meaning and pronunciation; convincing facial motion cannot fix a wrong script.

Continue with an approved result

Use an approved result in the planned video sequence. For another language, decide whether you need new reviewed narration or the separate Dubbing workflow, and inspect timing and meaning again.

Model controls to review

• Kling Avatar

  Check portrait, audio and duration requirements before preparing the inputs.

Related workflows

• narration

  generate spoken audio from text

• record speech

  capture a recording for voice workflows

• multilingual narration

  create speech in supported languages

• video dubbing

  translate spoken video into other languages

Sources and basis

• Kling Avatar inputs and settings in DayGen — DayGen product information; accessed September 20, 2026