How to make a faceless video for TikTok

Last updated 2026-08-16 · 5 min read

What is the short version?

Faceless channels live or die on consistency. Every episode has to sound and look like the last one, which is the part that gets tedious around episode nine and is exactly the part worth handing over. The topic is your job; the sameness is not.

1. Go to Vizard Agent and say the topic in a sentence or two.
2. Say the length, the feed, and the register you want.
3. From the second episode on, say "same as last time" and let Vizard Agent match it.

What do you need before you start?

A topic with something specific in it, and a decision about register. Vizard Agent writes the script itself, so what it needs from you is the fact worth building thirty seconds around and the tone that fact should be delivered in — those two choices set everything downstream.

• A topic with a fact in it. "The immortal jellyfish can reverse its own ageing" carries a video. "Ocean facts" does not.
• A length. Thirty seconds is the usual ask and it forces the script to stay tight.
• A register. Calm documentary, urgent, eerie. This decides the voice and the pacing more than the words do.
• Your previous episode, if there is one. Same project, or upload it. This is what makes episode ten look like episode one.

What do you type into Vizard Agent?

Two clauses do the work: the fact, and the instruction to match what came before. The second clause is the one that turns a series into a series rather than ten separate videos that happen to share a subject.

Prompt

Variants worth knowing:

• First episode. Drop the last clause and describe the look instead: "clean documentary, white bold captions, no music bed." Vizard Agent will then have something to match against next time.
• Your own voice. Upload a voice sample and ask Vizard Agent to narrate in it.
• A batch. Ask for one, approve the style, then ask for the rest. Approving a look once beats correcting it five times.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real episode of an existing channel. The first half of the run is not about this episode at all — it is about the previous one, which is the whole reason the series holds together.

1. Checks the project's history for style consistency.
2. Analyses the previous final video's voiceover and structure.
3. Extracts frames from that previous video and looks at them one at a time to read the fonts and caption style.
4. Listens to the previous voiceover and music to characterise the voice and the bed underneath it.
5. Checks which caption styles and presets are available, then matches rather than invents.
6. Writes, narrates and builds this episode against what it has just measured.

Steps 1 to 4 are not Vizard Agent remembering a preference. They are Vizard Agent going and looking at the last episode, which is why the match holds even when your own memory of the settings does not.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 30fps, 43.75 seconds, AAC audio. It is ready for the feed exactly as delivered, with the captions burned into the picture and the narration already mixed against the music bed, so there is no assembly step waiting for you at the other end.

If you need a different length or a second cut for another platform, say so in the same conversation. Vizard Agent still holds the script, the narration and the visuals, so a 60-second version is a re-cut of material that already exists rather than a fresh build, and it costs accordingly.

Timing was not measured separately for this kind of job. The closest measured work runs a median of 28 to 38 minutes end to end, with the middle half spread considerably wider. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

The failure mode here is not technical. It is that a faceless video is only as good as the claim at its centre, and Vizard Agent will write a confident script around a weak claim just as readily as a strong one.

• A topic with no facts gives a script with no spine. Bring the specific number, the strange detail, the thing that sounds untrue. That is what the video is built around.
• The claims are yours once you post them. If accuracy matters — health, law, money — read the script before it goes out. Vizard Agent writes what the topic implies, and it is your channel that carries it.
• The first episode sets everything. Getting the look right on episode one is worth an extra round, because every later episode is matched to it.
• Generated visuals are generic by default. If you want a specific identity, describe it, or supply images for Vizard Agent to work from.
• Series drift is real over months. Matching works off the episode you point at, so if episode forty is matched to thirty-nine and thirty-nine to thirty-eight, small changes accumulate. Point Vizard Agent back at an early episode occasionally to reset the reference.
• Thirty seconds holds about one idea. Two ideas in thirty seconds is where faceless videos start to feel like a list being read out.

How do you fix a result that came back wrong?

Say what is off and Vizard Agent adjusts that layer only. The script, the narration, the visuals and the captions are separate pieces of the same project, so changing the voice does not rebuild the pictures and rewriting one line does not re-narrate the whole thing.

• "The voice is too fast." Re-narrates at the pace you name.
• "Change the second visual." Swaps one shot.
• "This line is wrong." Rewrites and re-narrates that line, leaving the rest.

How does Vizard Agent compare to doing it yourself?

The manual version of a faceless channel is a script in one tab, a text-to-speech tool in another, stock or generated visuals in a third, then an editor to assemble and caption it. None of it is hard. All of it is repeated every single episode, and the consistency has to be maintained by hand.

Common questions

Does it write the script too? Yes. Give Vizard Agent the topic and it writes, narrates and builds. You can also paste your own script if you would rather keep that part.

Can it use my voice? Yes. Upload a clean sample and ask Vizard Agent to narrate in it.

How does it match my previous episodes? By opening them. Vizard Agent reads the fonts and captions off the frames and characterises the voice and music from the audio.

Can I give it my own script? Yes. Paste it and Vizard Agent narrates and builds from your words instead of writing its own, which is the usual arrangement once a channel has a voice of its own.

Can it do a whole week at once? Ask for one first and approve the style. After that each episode is a short prompt, because the reference already exists.