How Heady Makes a Video: From a Topic to a Finished, Chaptered Upload

People ask what happens between typing a topic into Heady and downloading a finished 12-minute video. This post walks through it, stage by stage, including the decisions we made and why.

The short version: Heady writes the script with you, builds one consistent illustrated world for the video, animates it without a video model, narrates it, times every word, adds music, checks the result, and hands you the file plus everything YouTube wants in the description.

1. A topic becomes an outline

You start with a topic and a length. Heady plans the video first: the question the video answers, and a list of chapters, each with its own smaller question.

2. The script, beat by beat, and you approve it

Heady then writes the script chapter by chapter. Each chapter is split into beats: a line or two of narration, plus a description of what is on screen while it is spoken.

Nothing is produced until you approve the script. This is deliberate. The script is where a video succeeds or fails, and it is far cheaper to fix a chapter title than a finished minute of video. Reorder chapters, rename them, or change the angle entirely.

We also think the approval step matters for the channel itself. YouTube's monetization policy excludes "AI-generated content made with generic or unoriginal templates giving the impression of mass production" (YouTube Help). A person deciding what each video says is the opposite of that.

You can edit any line. Change a sentence that does not sound like your channel, fix a fact, add the detail only you know. The script follows the structure we describe in How to structure a long YouTube script for retention: a cold open with a question, chapters that each answer something, and a payoff at the end.

3. One world per video

Image models are good at single pictures and bad at sequences. Ask one for 120 images of the same place, and you get 120 slightly different places: the building moves, the colours shift, the character changes face.

Heady handles this by never asking for a fresh picture after the first one. The first frame of a video sets the look: style, palette, places, characters. Every later frame is an edit of the first, changing only what the beat needs: the time of day, what is happening, what has been added or removed. The result is one consistent world across the whole video, so viewers watch a story unfold in one place instead of a slideshow of unrelated images.

4. Motion without a video model

A still image held for ten seconds feels dead. Generating video for every beat is the obvious fix, and an expensive one. At published 2026 rates of $0.05 to $0.10 per second for video models such as Veo 3.1 Lite and Fast, ten minutes of generated footage costs roughly $30 to $60 in model fees alone (dev.to, ModelsLab). Generated clips also drift from each other, which undoes the consistent world from step 3.

Heady animates the stills instead:

• Depth parallax. For every frame, Heady estimates a depth map: how far each part of the picture is from the camera. That splits the image into layers, and a virtual camera moves through them. Near objects move more than far ones, which reads as real depth.
• Camera moves. Slow pushes, pans and punch-ins on the same frame give a single image several shots' worth of life.
• Smooth transitions. Because consecutive frames share the same world and camera, Heady can move between them smoothly: computing the in-between frames, or bringing the next frame in by depth, far layers first, while the camera keeps moving.

The result is a video that moves continuously, at the cost of still images.

5. Motion mode: graphics built as code

Some beats are not scenes. A number that grows, a chart that compares two things, a timeline, a map with a route, a title. Image models are poor at these: they misspell words and invent numbers.

In Motion mode, Heady builds these beats as animated motion graphics written as code: charts, kinetic titles, timelines, counters. The text on screen is real text, so it is spelled correctly and the numbers are the numbers in your script. Motion scenes need no image generation at all.

You can make a whole video in Motion mode, which suits data-heavy explainers, or mix it with illustrated beats.

6. The voice, and every word timed

A natural AI voice reads the approved script. Pick one of Heady's voices and keep it across uploads, so your channel has one recognisable narrator.

After the narration is generated, a speech recognition pass on our servers listens to the audio and finds when each word is spoken. The subtitle text comes from your script, so names and terms are spelled the way you wrote them; only the timing comes from the audio. That is what makes word-by-word subtitles possible: each word lights up as it is spoken.

The same timings set the video's rhythm. Each beat lasts as long as its narration, so pictures change with the words rather than on a fixed clock.

7. Music, ducked

A music bed runs under the narration. It is ducked: it drops while the voice speaks and comes back up in pauses and on chapter cards, so it adds atmosphere without competing with the words.

8. Assembly: chapters you can see

For long-form, the video is assembled at 16:9 with the pieces that make 20 minutes navigable:

• Chapter cards at the start of each chapter, using the titles from your outline.
• Lower-thirds for names, places, dates and numbers.
• A progress bar showing the chapters, so viewers know where they are.
• Word-by-word subtitles in sync with the voice.

Shorts are assembled at 9:16, with the same voice, subtitles and look.

9. Quality checks

Before a video reaches you, Heady checks it. The checks look at things that are easy to miss when you are close to a project: the length against the plan, black frames, loudness, how clearly the speech comes through, and on-screen text that is too long to read in the time it is shown. Problems are flagged so they can be fixed before you publish.

10. What you download

For a long-form video:

• The finished MP4.
• YouTube chapter timestamps, starting at 00:00, ready to paste. They come from the finished video, so each one matches where its chapter actually starts. Our chapters guide explains why that matters.
• An SRT subtitle file, so YouTube's captions match your script.
• A description to paste or edit.

You review the finished video and publish it yourself.

An example, start to finish

Say you type "What if the Moon disappeared?" and ask for about 12 minutes. The outline Heady drafts might look like this:

Question: What would actually change on Earth, and how fast?

00  The first thing you would notice
01  The tides shrink, but do not stop
02  Nights without moonlight
03  The animals that keep time by the Moon
04  A calendar without a month
05  The slow wobble
06  Could anything replace it?
    Payoff: which change hits first, and which takes longest

Heady writes the script around it, about 1,800 words at a typical narration pace. You might swap chapters 3 and 4, rename chapter 5 to something clearer, fix two lines, and approve. The finished video has a progress bar with seven segments, a chapter card at the start of each chapter after the cold open (six cards), and seven timestamps in the description, all with the titles you approved. It uses 12 credits.

What Heady does not do

It is worth being clear about the limits:

• No video model. If you want footage of people moving, talking or acting, Heady is not the tool.
• No stock footage and no avatars. Every visual is generated for your video, as an illustrated still or a Motion graphic.
• English first. Scripts and voices are in English at launch.
• It does not choose your topics. You bring the topic and approve the script. That is the part that makes a channel yours.

What it costs

Heady is priced in credits, and 1 credit is 1 finished minute of video. A 12-minute video uses 12 credits. Plans and credit packs are on the pricing page.

If you are starting a channel from scratch, read How to start a faceless YouTube channel first. If you already have long videos, One long video, ten Shorts shows how to get more out of each one.