How to make a talking-head video in Affogato, from script to episode

A talking-head video is easier to watch when the speaker has one clear thing to say. You do not need a long introduction explaining why the subject matters. Start with the useful point, explain it in words someone would actually say, and stop when the viewer has what they need.

We created a new host, Julian, for this walkthrough. He gives a short lesson about checking product images. The example covers the complete process in Affogato: creating and saving the host, choosing a set and voice, listening to a table read, generating the performance, and reviewing the assembled episode in Video Editor.

Julian on the Creator desk set, generated as the starting image for the episode.

Write the script for someone listening

Read the script aloud before generating anything. A sentence that looks tidy on a page can be awkward to speak. Replace vague language with a detail the listener can picture, cut repeated phrases, and keep the sentence length comfortable.

Our script teaches one review habit:

Before you generate a product campaign, choose the details that must stay unchanged. Check the shape, the cap, and the lettering in every result. A beautiful photograph is only useful when the customer can still recognize the product.

The advice is concrete enough to demonstrate with supporting product images later. It does not invent a personal experience or promise a result. For your own video, decide whether the host is teaching, introducing a product, answering a question, or explaining a process before choosing the visual treatment.

If the script covers several ideas, divide it into short sections. In Talking Head Studio, a blank line creates a new segment or cut. Use that deliberately. A paragraph break should correspond to a sensible change in the explanation, not an accidental line break pasted from a document.

Create a host whose face is easy to animate

Julian’s portrait was generated in Affogato Image Studio and saved as a new character for this guide.

Julian has swept salt-and-pepper hair, a clean-shaven face, and a simple navy top. We kept the portrait clear and the mouth unobscured. Dramatic shadows, hair over the face, or a hand across the chin can make an attractive photograph less useful as a speaking reference.

You can generate the portrait in Image Studio, then save it through the character setup flow. We named this character Julian — Talking Head Guide. Reusing a saved host makes it easier to build a series without inventing a new presenter for each episode.

The Inspiration gallery is useful when choosing wardrobe, light, and background. Open a reference, inspect the Image prompt, and adapt it through Copy or Use Inspiration. Choose details that make the speaker easier to watch. A room full of decorative objects can compete with a short explanation.

Set up the episode in Talking Head Studio

1. Open Talking Head Studio and enter the script.
2. Select Julian — Talking Head Guide under Host.
3. Choose Vertical, 9:16, for this example.
4. Select an engine. We used Kling AI Avatar v2; the current studio shows estimates beside the available engines.
5. Choose Creator desk under Set and Waist-up framing.
6. Under Voice source, choose Generated voice. Open the voice picker and select Roger, the American male voice used here.
7. Select a 1080p export.

The script and new host in the actual Talking Head Studio setup.

The selected set, framing, Roger voice, export resolution, and generation controls.

Choose the voice by listening, not just by reading its description. The voice should suit the pace and subject. A highly dramatic delivery can make a simple instruction sound unnatural. You can also use your own audio when that is the right source for the episode.

Listen to the Table read before rendering the face

Select Table read to hear the script in the chosen voice. Check pronunciation, pauses, and whether the important words are clear. Listen to the beginning and end as carefully as the middle. A clipped first word or a rushed final sentence can spoil a short video.

Change awkward wording at this stage. It is much easier to hear that a sentence needs rewriting before paying to animate it. The preview is a production step with its own estimate; review the current credit information rather than assuming a preview means the whole episode is free.

Our episode reused the previewed speech when generating the performance. Keep the page open while the episode runs so the studio can carry the sequence through to assembly.

Generate the complete performance and review the result

Select Generate episode. The studio prepares the host on the set, produces the speech and speaking performance, and assembles the episode. Wait for Episode ready. A finished audio preview or a completed keyframe is not yet the final video.

Julian’s completed episode, about 14.7 seconds long. The animated preview is silent; open the MP4 to hear Roger’s narration. Watch the full video (MP4).

Watch the mouth against the spoken words. Check the eyes, teeth, and face during movement, then look at the hands and clothing. Listen for pronunciation errors, unexpected sounds, or a cut that arrives before the sentence finishes. We also checked the exported audio against the script so the article describes the finished performance, not just the text entered into the form.

A short, single-point episode is a useful first test because problems are easier to identify. If the voice is wrong, change the voice or script. If the set is too busy, simplify the set. Changing everything at once makes it harder to understand what improved the result.

Open the assembled episode in Video Editor

Use Polish in the editor to open the episode’s project. In this example the timeline contains the completed spoken clip with its audio. You can trim a pause, add supporting footage, or add a title where it helps the viewer.

The completed Julian episode in the real Video Editor timeline.

For this script, useful supporting footage would show a product’s shape, cap, and lettering at the points where Julian names them. Keep the spoken explanation clear if you add music. If you add captions, read them through and correct product names or specialist terms before exporting.

Review the exported file at the size and orientation people will actually watch. A face that looks clear on a large monitor may be too small on a phone, and a title that fits the editor can still sit under a platform’s interface. Save the script, host, voice, and set choices with the episode so the next lesson starts from an established presentation.