How to introduce a new character into a video series you already run

Last updated 2026-08-24 · 7 min read

What is the short version?

Adding a character to a running series is not a standalone video. It has to look like the series, sound like the series and sit next to the last episode without anyone noticing the join, which means the existing episodes are the specification.

1. Go to Vizard Agent and give it an episode you have already published.
2. Describe the new character and what they are there to do.
3. Say the length and where the video will run.

What do you need before you start?

An existing episode and a description. Vizard Agent derives the visual style, the subtitle treatment and the pacing from what you have already made, so you do not need to write a brand brief, and the character description is the only genuinely new information in the job.

• An existing episode. Ideally the most recent one.
• The new character. Who they are and what they represent.
• Any existing characters. So the two can appear together.
• Your logo or brand colours. For the captions and the closing frame.
• The length. Ten to fifteen seconds for an introduction.

What do you type into Vizard Agent?

Point at the reference and describe the role. Vizard Agent uses the existing episode for everything about how it should look and sound, so your instruction only needs to carry who the new character is and what job they do in the series.

Prompt

Variants worth knowing:

• A fuller shot. A full-body introduction reads differently from a close one.
• A specific voice. Warm and mature, young and bright; it defines the character.
• Brand-coloured captions. They tie the new character to the existing series.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real character introduction. The whole first third is study — the reference video, the previous project's images, the earlier deliverables — before a single new frame is generated.

1. Inspects the reference video and the past project details.
2. Extracts the reference frames and tiles them to read the visual style.
3. Downloads the project's reference images and builds a contact sheet of the existing characters.
4. Extracts the reference audio and its word timings to learn the pacing.
5. Uploads the reference images and audio clips so the new generation has something to match.
6. Generates the new character alongside the existing one, walking and talking in the established setting.
7. Reviews the generated frames, measures the loudness and sources matching music.
8. Renders two subtitle styles as test cuts and compares them on frames against the previous deliverable's look.
9. Searches for a mature, warm voice for the new character, generates two candidates and compares them, regenerates the whole video as a full-body shot with the chosen voice, then re-renders the captions in the brand's blue with a white ending.

Step 9 is the honest shape of this work. The first version was good and the second was right: a fuller shot introduced the character better, and finding that out meant building it rather than debating it.

What does the result look like?

From the run this page is written from, probed on the delivered file: 720x1280, H.264, 24fps, 11.8 seconds, AAC audio. Vertical, under twelve seconds, the new character introduced alongside the established one in the series' own setting, with brand-coloured animated subtitles and a matching closing frame.

Twelve seconds is short for an introduction and correct for a series. The audience already knows the world; all this video has to do is add one person to it and say what they are for.

When does this not work well?

A series builds up recognition slowly over many episodes, and a single mismatched one spends a good deal of that credit. Vizard Agent matches deliberately against what you have already published, and some kinds of consistency turn out to be much harder to hold than others.

• A generated character will drift. Each new episode is another chance for the look to move.
• Health and medical claims are regulated. A friendly character does not soften what the video says.
• The reference sets the ceiling. If the existing episodes are inconsistent, there is no single style to match.
• Two characters in one shot is harder. Generation handles one figure more reliably than two.
• Twelve seconds introduces, it does not explain. The character's first real topic is a separate video.

How do you fix a result that came back wrong?

Name what does not match. Vizard Agent keeps the reference analysis, the generated takes, both voice candidates, the subtitle style tests and the previous deliverable's frames, so a change of voice or a caption restyle is a re-render rather than a rebuild.

• "The voice is too young for him." The other candidate, already generated and compared.
• "He should be full-body." Regenerated at that framing, as this run did.
• "The captions are the wrong blue." Re-rendered in the exact brand value.

How does Vizard Agent compare to doing it yourself?

By hand this means opening your last episode, trying to remember which font and which blue you used, and then generating a character who looks like they belong to a different show. Matching a series from memory is the part that quietly fails, because nobody notices the drift until three episodes later.

Common questions

Why does Vizard Agent study the old episode so closely? Because the existing series is the specification. Vizard Agent reads the frames, the audio pacing and the subtitle treatment out of it, which is more reliable than any written brand brief and catches the details nobody wrote down.

Do I need to send an existing episode? Yes, and it does most of the briefing. Vizard Agent reads the style, pacing and captions from it.

Can the new character appear with the old one? Yes. Vizard Agent generated both of them together, walking and talking in the series' established setting, on this run.

How long should an introduction be? Ten to fifteen seconds. Vizard Agent builds it to introduce the character, not to teach a topic.

Can I choose the voice? Yes. Describe the character and Vizard Agent generates candidates to compare before committing.

Will the captions match our brand? Yes, down to the exact colour value if you supply it.

Why regenerate the whole video for a framing change? Because the framing changes the performance. A full-body shot introduces a character differently from a close one, and Vizard Agent rebuilt rather than cropping the existing take.

Can it keep the character across future episodes? Yes, by matching each new one against the last. Over a long run the look drifts and needs re-anchoring.

Does Vizard Agent check the finished cut? Yes. It proves the audio balance and reviews master frames before delivery.