AI images and video for faceless YouTube

20 video segments5 creators1 h of video → 6-minute read

Updated 14 November 2026.

Make the images first and animate them second. That order is what keeps characters and style the same from scene to scene, and consistency is the hardest part of AI visuals. Fix the description of each character, use your first good image as the style reference for the rest, and keep motion subtle. The tools change every few months; this order of work does not.

The clips show specific tools, and several creators promote the tool they use. Menu names will have moved by the time you read this, so follow the method, not the buttons.

Segment 1 of 20 · starts at 4:00
Scene prompts with fixed character descriptions for consistency
Wokobucci coach · 4:00 – 7:30 · uploaded Nov 2025

Plays from YouTube when you press play. Watch on YouTube

Step1

Consistent characters across scenes

Fix the description. Before any image is made, have the chatbot write a fixed description of each character: age, build, clothing, colours. Repeat that exact description in every scene prompt, and generate the scenes one at a time in the same chat. Wokobucci coach 4:00

Save the character. Tools with a character feature let you design a figure once and attach it to every scene. Joshua Mayo creates a fox this way, describes the three beats of the story in a chat, and receives one continuous video with visuals, narration and music. One long generation keeps the character and the world more consistent than many separate ones, and single shots can be adjusted on a timeline afterwards. Joshua Mayo 9:25

Make a character sheet. From one good image of a character, generate a reference sheet: the same figure from several angles, with close-ups of the face, in one picture. Then attach that sheet as a reference to every image and every clip you generate, even when you already have a start frame. Two sheets together put two consistent characters in one scene. This needs a generator that accepts reference images, not only image-to-video. AI Video School 0:00

Dan Kieft builds a whole animated story this way. He creates the characters first, all in one visual style, and a sheet for each before any scene exists. Dan Kieft 1:15

For realistic people, build the character from several reference photos and reuse the resulting images as the base for every clip. Joshua Mayo 4:38

Use only photos you have the right to use: yourself, a consenting person, or a generated face. A real person shown doing something they did not do has to be disclosed at upload and can break YouTube's rules.

Step2

A style-matched image series

  1. Define one visual theme before generating: for example black and white with one subject in the centre.
  2. One prompt per story beat. Have the chatbot turn the script into a list of image prompts that follow the story.
  3. Generate options and choose. Several per prompt.
  4. Use your best image as the style reference for all the others, at low strength, so they share a look without becoming copies.
  5. Upscale the ones you keep.

AI Guy 2:50 · AI Guy 8:00

Storyboard with images before any video. Images are cheap and video is not, so make the mistakes here. Dan Kieft makes one image per scene from the character sheet and a prompt that names the character, the action, the light and the style, and lays them out in story order. When he left the style line out, he got cartoon figures among realistic people. Dan Kieft 3:00

Props work the same way: generate the photo or the flyer as its own image and refer to it in the prompt for the scene. Some tools can also turn one finished image into several camera angles. Dan Kieft 6:20

A simpler routine for story videos: paste each line of the script into the image tool as a prompt, and mark the finished lines in the script document so you can see what is left. Which source fits depends on the format: stories lean on AI images, list videos on stock footage and photos. Joshua Mayo 20:22

State the format in the prompt: landscape for normal videos, portrait for Shorts. Some tools default to portrait. Wokobucci coach 4:00

Step3

Animate stills into clips

Upload the image as the first frame of an image-to-video tool and reuse the scene prompt that produced it. Work through the scenes in order and keep the files in story order. Animating a finished image keeps the characters the same across clips; generating video straight from text does not. Wokobucci coach 7:30

Keep the motion setting low to medium. A slow push-in or a light movement looks natural, and strong motion distorts faces and hands. AI Guy 2:50

If the characters speak, each generated clip gives them a slightly different voice. Run every clip of one character through a voice changer with the same target voice. Dan Kieft 14:15

In the editor the clips follow the storyboard order. When lip movement lags behind the sound, move that clip's audio a few frames earlier. Sound effects and music come last. Dan Kieft 16:20

Step4

Prompt and camera craft

  • Short prompts. Describe what happens, the mood and the atmosphere. Leave the camera to a preset.
  • Camera presets. Dolly zooms, crash zooms and drone moves are ready-made. Subtle ones suit realistic long videos, and stronger ones suit Shorts that need to catch the eye.

Joshua Mayo 7:28

For a shot that has to go from one exact picture to another, generate both images and use them as the start and end frame, with a prompt that describes what happens in between. Reference the same character image in both to keep the person the same. Joshua Mayo 13:08

You often have only one of the two frames. Make the other by editing the image you have: the same room with the door closed and nobody in it becomes the start frame for a clip of someone walking in. Dan Kieft 8:25

To make two clips play as one take, save the last frame of the first clip and use it as the start frame of the next, and describe the camera move between them. Dan Kieft 10:10

A multi-shot mode gives several camera angles in one generation, with one short prompt per shot. Generating the shots separately gives more control. Dan Kieft 11:35

Step5

One-click templates and cinematic controls

Tool-specific; the tool reviews go into more detail.

Some platforms sit on top of several video models and let you switch per shot. Joshua Mayo presents Higgsfield this way and says the default model is fine to begin with. Joshua Mayo 1:21

Two ends of the same tool:

  • Templates need an image or two and return a finished clip, such as a transition between a first and a last image. They are quick and recognisable, so use them sparingly. Joshua Mayo 11:01
  • Full control lets you choose a camera body, lens and focal length and combine camera moves. Worth learning only when a specific look matters. Joshua Mayo 13:08

A reminder from the policy lesson: a video that is one template applied again and again is what YouTube calls inauthentic content. The visuals are one of five pillars. Your script, your choice of what to show and your edit are what make the video yours.

Checklist

Your ticks are saved in this browser only.

← PreviousChoosing a voice for your faceless channel: AI, your own or hiredMark done & continue →Faceless video formats: stories, animation, explainers and documentary

The full video on YouTube