How do you make a talking-head video with an AI character and voice?

Choose a saved host, write a short script, select a voice, and generate the speaking video. Review the narration before rendering the full result, then check the mouth movement and facial consistency throughout the clip.

Affogato has a dedicated Talking Head Studio. It must be enabled for your account, along with the models and supporting services it needs. If the page says it is not switched on, use the alternative lipsync workflow below when available.

We'll plan a short morning-routine explainer. It uses a fictional host rather than a real person's unapproved likeness or voice.

1. Prepare an unobstructed host

Create or choose an existing character. A clear face, ordinary expression, and simple lighting make the first test easier to assess. Avoid hands across the mouth, hair covering the lips, or extreme head angles.

Supplied host-reference candidate. This is a still portrait, not a rendered talking-head video or evidence of lip-sync quality.

Use the example to plan framing: the face is large enough to read, with room around the head and shoulders.

2. Write a script that sounds spoken

Open Talking Head Studio and select the Host. Use short sentences in your script. For example:

Before the day gets busy, take a moment to settle in. Open the curtains, put your everyday essentials within reach, and choose one thing you want to do first. A small routine can give your morning a little structure.

Read the script aloud. Remove phrases you would not naturally say. Spell unfamiliar names in a way your chosen voice can pronounce, and listen before deciding the wording is final.

3. Choose the set, framing, and engine

Choose a Set, or upload your own set image if you want a particular background. Select framing that keeps the host prominent. Use the available Engine choices and their displayed prices to decide what to test.

Start with a short script and one setup. The engine's displayed rate helps compare options; the overall run can include more than the speaking-video stage. Review the full estimate before generation.

4. Choose and audition a voice

Under Voice source, choose Generated voice and open the voice library. Listen to available free samples to find a suitable accent and delivery. A saved custom voice may also appear where the supporting model is enabled.

Run the table read to hear your script before generating the episode. Generating your script's speech can consume credits; a library sample and a script-specific table read are not the same action.

If you already have an authorized narration recording, choose the uploaded-audio route instead. Listen for background noise and abrupt cuts before using it.

5. Generate the episode

After approving the speech and setup, generate the episode and keep the page open while it progresses. The browser coordinates several stages: visual setup, speech where needed, speaking segments, and assembly.

Longer scripts may become multiple segments. A successful narration preview does not mean the visual stages are finished. Wait for the completed output before reviewing the assembled video.

6. Check speech and face together

Watch once with sound, then watch the mouth closely. Check whether the lips match obvious speech sounds, the eyes remain stable, and the face stays recognizable. Listen for missing words or awkward pauses at segment boundaries.

If one sentence sounds wrong, correct that sentence and preview again. In the same session, the workflow can reuse some completed stages, but do not assume every revision is free or survives a page reload.

Alternative: lipsync an existing video

Open Video and select Lipsync. Supply a suitable source video and an audio track. The audio slot supports upload and, when the necessary voice models are available, Generate voiceover.

The source input is a video, not merely a portrait. Use a suitable authorized clip or create an approved motion clip first. Then generate and review the synchronized result.

Can you automatically translate the video into every language?

This tutorial does not assume automatic translation. Prepare the translated script separately, choose a compatible voice, and have a fluent speaker review pronunciation and meaning before using it.

Next step: make one short explainer in Talking Head Studio, then reuse the workflow for a creator-style ad.