Browse documentation

Setup faceless

Build your cartoon avatar and a faceless style, so your shorts use only your voice and never your face.

setup-faceless creates the avatar that plays you on screen. You record only your voice, and the avatar acts each sentence in a small frame at the bottom, over dense motion design. Your real face never appears.

It takes about 20 minutes, mostly image generation. You do it once: after that, any take can become a faceless short.

Invoke

Terminal
/setup-faceless
/setup-faceless pose he holds a cup of coffee
/setup-faceless poses
/setup-faceless design
/setup-faceless cta
/setup-faceless preview

The first form runs the full onboarding. The others jump to one part: a new pose, the missing poses, a new design, the profile picture, or the test video.

What it does

  1. Checks your main profile (name, CTA, captions, music). Without one, it suggests running studio-setup first.
  2. Checks image generation: Codex if you are logged in (recommended, it keeps the character most consistent), otherwise a Gemini or OpenAI key.
  3. Asks 5 short questions: the style name (faceless by default), what the avatar is based on (1 to 3 photos, a drawing you already have, or a description), and the outfit it always wears.
  4. Draws the same character in 4 art styles (marker, flat vector, adult animation, comic) on one board. You pick the one that looks like you, or describe your own style. Not convinced: it redraws with your note.
  5. Draws the poses while you wait: the core pack (14 poses: talking, pointing, shocked, doubt, shrug, laugh, thumbs up, wave, dance) or the full pack (27). Each pose is checked for the same face, the same outfit and a whole body in frame, and bad ones are redrawn.
  6. Makes the CTA picture from the avatar's head, used on the end card and as a profile picture.
  7. Writes your faceless style in .short-editing/styles/<name>/: audio only, the avatar small at the bottom, teleporting and acting each beat, never your real face.
  8. Renders a test video of about 20 s with your avatar, in about a minute. You say perfect, or redo some poses, the picture or the whole design.

A run that stops halfway keeps what is done: run the same command again and it only draws what is missing. Your photos stay private in avatar/photos/ and never appear in a video.

Use it

Record any take, since only the voice is used: a phone or webcam video, a YOLO take in the Recorder, or an audio file. Then:

Terminal
/editing --style faceless latest

Or just say "make a faceless short from my last take". A pose is missing in a short? Run /setup-faceless pose <what it does>: the new pose is drawn in the same style and is available in every later short.

What you need

  • Codex logged in (npm i -g @openai/codex, then codex login, free with a ChatGPT plan), or a Gemini or OpenAI API key pasted in the .env file the skill opens for you (never in the chat).
  • 1 to 3 clear photos, a drawing, or a description of the character.
  • Patience for image generation: designs take 2 to 3 min, the core pack about 8 min, the full pack about 15 min.

Next