Guides

How to make faceless Reels and TikToks with AI

· 7 min read

Get the iPhone app

Dora is made by Saga Labs, the publisher of this site.

Short answer: a faceless Reel or TikTok has four layers: a script, a voice, visuals and captions. AI can produce each one, so you never film yourself: an assistant writes the script, a text-to-speech voice reads it, an AI video model makes vertical clips for each line, and an editor adds captions and music. Before you post, label realistic AI content: TikTok and YouTube both require it.

The four layers, and tools for each

Layer What it does Tools people use
Script The hook and the story, line by line ChatGPT, Claude, Gemini
Voice Narration without recording yourself Text to speech in CapCut, ElevenLabs, Dora on the web; or your own voice
Visuals Footage for each line, with no camera AI video apps such as Dora, Kling AI or Runway; AI stills in sequence; stock footage
Captions and edit Subtitles, cuts, music CapCut (auto captions), Captions (subtitles in 90+ languages, per its listing)

If you want a presenter without being on camera, there is a fifth route: an AI avatar. Captions offers an "AI Twin" from your selfie and AI actors, and HeyGen makes avatar videos with voices and translations, according to their listings.

The workflow, step by step

  1. Pick one format and stick to it. Common faceless formats: atmospheric b-roll behind a text hook; a narrated story; a slideshow of AI stills; a recurring character who is not a real person. A series is easier to grow than random one-offs; see how to turn a story script into an AI video series.
  2. Write the script with an assistant. Ask for a hook in the first line, one idea per video, and the script split into short lines, each with a description of the shot to show. Those shot descriptions become your video prompts.
  3. Make the voiceover. Paste the script into a text-to-speech tool and pick one voice for the whole account, so it becomes recognisable. Or record it yourself: faceless does not have to mean voiceless.
  4. Generate the visuals, vertical. Set every clip to 9:16. One prompt per line of script, one movement per clip ("slow aerial over a misty forest at sunrise"). For a recurring character, give the model the same reference photos every time.
  5. Assemble and caption. Put the clips on the voice track in an editor such as CapCut, add auto captions (most viewers watch without sound) and music.
  6. Label it and post it yourself. See the rules below.

Label realistic AI content: what the platforms say

  • TikTok: "We require creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people" (Community Guidelines, Integrity and Authenticity). Unlabelled content "may be removed, restricted, or labeled" by TikTok, depending on the harm it could cause, and TikTok does not allow AI content that misleads about matters of public importance or harms individuals.
  • YouTube (Shorts included): creators must disclose realistic AI content, for example content that makes a real person appear to say or do something they did not do, alters footage of a real event or place, or generates a realistic scene that did not happen. You do it with the AI use setting when you upload.

Clearly unrealistic content, such as cartoon styles or obvious fantasy, generally does not need the label on YouTube; for anything that could pass as real, label it.

What Dora gives you for faceless video

Dora is one way to make the visuals, and on the web it also covers the voice. This is what it offers today, read from its code on October 23, 2026 (the web app's model list and the MCP server that ChatGPT and Claude use):

Need What is in the code
Vertical clips SeeDance, Grok, Veo and Gemini Omni take 9:16 directly (the SeeDance models also 16:9, 3:4, 4:3, 1:1 and 21:9); the Kling models follow the shape of your photo
Clip length Longest clip per model: Grok Imagine 15 s, SeeDance 1.5 Pro and SeeDance 2 12 s, Kling 2.6 10 s, Veo 3 Fast a fixed 8 s
Voiceover (web) ElevenLabs Multilingual v2 text to speech with 6 voices: Liam, Laura, Brian, Bella, James and Jessica
Sound effects (web) ElevenLabs SFX v2, sound effects from a text description
A talking character without filming (web) Lip Sync: one photo plus your audio, up to 15 seconds, 480p or 720p
The same character every episode Same Character: up to 7 photos of one character placed in any scene you describe

How we got this: model options were read from web/lib/models.ts and functions/src/mcp/tools.js in Dora's repository; the 15-second Lip Sync limit and the six voices are in the same files. In a timed test the same day, a 4-second 480p vertical clip was ready 30 seconds after the request (details).

What Dora does not do: it does not write captions into the video, edit clips together, or publish or schedule posts. You finish in an editor and upload from your own accounts. Dora is in the App Store and at doravideo.com; the voice, sound effect and Lip Sync tools are on the web app.

Mistakes that sink faceless accounts

  • Horizontal clips. A 16:9 clip in a 9:16 feed looks like a re-upload. Generate vertical from the start.
  • A different voice every video. The voice is the face of a faceless account.
  • No captions. Many people watch muted.
  • Real people made to say things. Both platforms require a label for this, and TikTok does not allow AI content that harms individuals. Use invented characters.
  • Too many ideas per video. One hook, one point, one payoff.

For tool-by-tool detail, see the best AI video apps for TikTok and Reels.

Frequently asked questions

What is a faceless TikTok or Reels account?

An account whose videos never show the creator's face. They are usually built from b-roll or generated footage, a voiceover (recorded or text to speech) and on-screen captions, around a single theme such as stories, facts, motivation or products.

Can AI make a whole faceless video?

Each part, yes: an assistant such as ChatGPT or Claude writes the script, a text-to-speech tool reads it, an AI video model generates vertical clips for each line, and an editor such as CapCut adds captions and music. Most people still assemble the final cut themselves.

Do I have to label AI videos on TikTok?

Yes when they look real. TikTok's Community Guidelines say it requires creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people. YouTube likewise requires disclosure of realistic AI content through the AI use setting when you upload.

What is the best aspect ratio for faceless Reels and TikToks?

Vertical 9:16. Generate clips in 9:16 from the start instead of cropping horizontal footage, so nothing important is cut off and the video fills the screen.

Can I make the voiceover with AI?

Yes. CapCut has text to speech, ElevenLabs is a dedicated voice tool, and Dora's web app includes ElevenLabs Multilingual v2 with six voices. Choose one voice and keep it for the whole account.

Does Dora post the videos for me?

No. Dora generates video, images, voice and sound effects and gives you the files. You add captions in an editor and post from your own TikTok or Instagram account, or use a separate scheduler.

Sources

Get the iPhone appUse it in Claude or ChatGPT

Keep reading

ComparisonsThe best AI video apps for TikTok and Reels (2026)Seven apps for making vertical AI videos for TikTok and Reels, ranked by what they actually do well, and how they charge.GuidesHow to make AI videos for TikTokVertical, scroll-stopping AI clips for TikTok — the right settings, models and a fast workflow.GuidesHow to turn a story script into an AI video seriesOne character that looks the same in every episode, one clip per scene, narration in your own language. The workflow, step by step.