Skip to main content
How-To Guides

How to Start a Faceless YouTube Channel With AI in 7 Steps

How to start a faceless YouTube channel with AI in 7 steps, and stay inside YouTube's 2026 rules on templated AI content, monetization and AI disclosure.

NKNick Koukoulakis27 min read
How to Start a Faceless YouTube Channel With AI in 7 Steps
How-To Guides
Pillar guide
AI Video Agents: The Complete 2026 Guide

Maybe you've tried a faceless YouTube channel with AI and ended up with 30 videos that all look the same. YouTube has a name for that: inauthentic content. Its monetization policy now names "AI-generated content made with generic or unoriginal templates" as content it won't pay for.

That doesn't make AI the problem. The same policy allows AI that serves an idea you bring. Synthesys has shipped AI voices since 2020 and AI avatars since 2021 (Synthesys). Here is how to start a faceless YouTube channel with AI that stays inside YouTube's monetization rules.

"Faceless video can carry a message, tell a story, teach or give a tip," says Peter Andrews, co-founder of Synthesys. "It has many uses, and it costs a lot less to make than filming."

A storyboard for a faceless YouTube explainer in 16:9, with timed scene cards, a narrator voice and an approve button

What is a faceless YouTube channel?

A faceless YouTube channel publishes videos without the creator on camera. A narrator's voice carries the story over footage, animation, graphics or screen recordings. AI can now draft the script, voice it and build the visuals. The topic choice, the angle and the final check still have to come from you.

Common faceless formats are narrated explainers, top-five lists, animated stories, deep dives and tutorials. The shape decides where a video runs on YouTube and how long it can be.

FormatLength limitShape
ShortsUp to 3 minutes (YouTube Help)Vertical
Long-form, unverified accountUp to 15 minutes (YouTube Help)16:9
Long-form, verified accountUp to 12 hours or 256 GB, whichever is less16:9

Verifying means confirming a code sent by text or voice call (YouTube Help). 16:9 is the shape for YouTube's main feed (Synthesys).

What do successful faceless channels look like?

The biggest faceless channels prove the format, and most of them weren't built with AI. They run on a narrator, animation or music instead of a face. AI changes the cost of making that kind of video, not what makes it watchable: a clear idea and a consistent look.

ChannelWhat replaces the faceSizeAI in production?
Kurzgesagt – In a NutshellFlat animation and one narrator25.3 million subscribers, 3.7 billion views (June 2026)None mentioned
HowToBasicHands-only visual gags, no voice17.8 million subscribers (2025)None mentioned
Lofi GirlOne looping animated character and 24/7 music15.8 million subscribersNone mentioned; drawn by an illustrator
Teddy Pooh (Instagram)An AI-generated teddy-bear-meets-poodle character100,000+ Instagram followers (The Hollywood Reporter, June 2026)Yes, fully AI-generated
Life Hack (Instagram)Illustrated science and everyday-tips posts, no presenter5.3 million followers, 424 postsNot stated on the page; its owner account teaches AI content creation

Sizes from the linked pages, checked 1 October 2026.

The Instagram account @adrianviralai, which links to Adrian Durso's YouTube channel, describes itself as "Owner @lifehack". Durso's Adrian Viral AI Marketing YouTube channel covers tools like Veo 3, Kling and Midjourney, with 46,300 subscribers and 249 videos since February 2025. He says he has grown 10 million followers organically (his community page); that figure is his own claim.

The fully AI side is under pressure, though. In June 2026 The Hollywood Reporter reported that YouTube has been demonetizing faceless channels as it cracks down on "AI slop" (The Hollywood Reporter). Creators told the paper the algorithm now favours videos with human faces, though YouTube hasn't said so. Craig Billings, known as Doctor NOS to his 1.7 million subscribers, put it bluntly. "The people who do the same content as me without their face in it, most of them are getting demonetized."

What these channels share is a recognisable look and a reason to watch the next video. That's the same thing YouTube's monetization policy asks for.

Which faceless formats does Synthesys have?

These are the faceless-friendly templates in the Synthesys app's template list, with the lengths the app allows (checked 1 October 2026).

TemplateStylesLength in the appDefault shape
Faceless storyStory, Listicle, Timeline, Facts, Biography, News RecapUp to 10 minutes9:16
Brainrot Edit (split-screen)Product Facts, Storytime, App Tips, Listicle Countdown, Hot TakeUp to 90 seconds9:16
Editorial explainer (Vox style)10 looks, from Editorial to Whiteboard Sketch and Infographic HUDUp to 5 minutes9:16
Whiteboard explainerOne whiteboard lookUp to 5 minutes16:9
Full YouTube videoPresenter, graphics or bothUp to 30 minutes (Synthesys)Any

Every format exports in 9:16, 16:9 or 1:1. Finished examples made by the agent, including an animation story and a Vox-style explainer, play on the Synthesys Video Agent page.

What is faceless YouTube automation, and is it allowed?

YouTube automation usually means running a channel where tools or freelancers handle most of the production. YouTube doesn't ban it. If you use automated tools or templates, its policy says the final video "must still demonstrate your creative vision". It also has to give viewers real value.

YouTube's monetization policy mentions automated tools once, in that line. What its reviewers check is the channel as a whole: its videos, its description, its titles and its video descriptions (YouTube channel monetization policies).

That's where automated channels tend to trip. Trouble starts when fifty videos come from one template, with only the topic swapped.

Which faceless YouTube niches work best with AI?

Niches where you can add your own explanation or story to every video: history, science, how-to, original fiction and explainers. The risky ones are expert advice on money or health, readings of other people's articles and clip compilations, because YouTube's policies treat them strictly.

NicheWorks asWhat to watch in YouTube's policy
History and science explainersLong-form, narratedRealistic AI recreations of real events or places need "AI use: Yes"
How-to and tutorialsLong-form, screen recording plus voiceLow risk: your own steps make each video different
Top-five and list videosLong-form or ShortsThe same structure every time can read as "templated storylines"
Original fiction and animated storiesShorts or long-formAllowed when it's "a unique character and narrative you invented"; the same storyline on repeat isn't
Personal finance, health, legalExplainersAn AI persona posing as a human expert can't be monetized
News and article readingsNarration over visualsReading material you didn't create, with no commentary, counts as reused content
Clip compilationsMontagesReused content unless you add a storyline and commentary

Policy wording from YouTube channel monetization policies and Disclosing use of GenAI content, checked 1 October 2026.

Can a faceless YouTube channel with AI get monetized?

Yes, if the channel meets the YouTube Partner Program thresholds and each video adds something of your own. Being faceless doesn't count against you. YouTube's policy targets AI videos made from generic templates that look mass-produced, and channels whose videos feel interchangeable.

To apply, you need 1,000 subscribers plus 4,000 public watch hours in the last 12 months. The other route is 1,000 subscribers plus 10 million Shorts views in the last 90 days (YouTube Help). Watch hours from the Shorts feed don't count toward the 4,000.

Hitting those numbers gets you a review, not a payout. YouTube says its monetization guidelines apply "to your channel overall", not just to single videos (YouTube channel monetization policies).

What YouTube's monetization policy allows and rules out for AI and template videos, quoted from the policy

The policy was renamed from "repetitious content" to "inauthentic content" on 15 July 2025. YouTube said the update clarifies that "repetitive or mass-produced" content was always ineligible. Its allowed examples include "using AI to visualize a unique character and narrative you invented".

One rule matters most for narrated channels. YouTube won't monetize channels that use AI personas presenting as human experts on health, legal, financial or political topics. Its examples include an AI "doctor" and AI podcast hosts giving investment tips.

"Take part in the process and make the edits yourself," Peter says. "With AI, I think there's nothing it can't do. But the final decision should always rest with a human."

How many views does a faceless channel need to get monetized?

On the long-form route, 4,000 watch hours is 240,000 minutes in 12 months. If viewers watch 4 minutes of each video on average, that's about 60,000 views. You also need 1,000 subscribers. The shorter the average watch, the more views it takes.

Average minutes watched per viewViews needed in 12 months
2 minutes120,000
3 minutes80,000
4 minutes60,000
6 minutes40,000
8 minutes30,000

This is our arithmetic from YouTube's thresholds, not a forecast. Only watch time on public long-form videos counts (YouTube Help). The Shorts route asks for 10 million Shorts views in 90 days instead.

The table makes one thing plain. Longer videos that people actually finish reach the threshold with far fewer views.

Do you have to disclose AI on a faceless channel?

Only when the AI content looks realistic. YouTube requires it when AI puts words or actions on a real person, alters a real event or place, or invents a realistic scene. AI help with scripts, captions, thumbnails or your own cloned voice doesn't need it.

Tick "AI use: Yes"No disclosure needed
AI-generated musicAn AI-drafted outline, script, title or thumbnail
AI footage of a real place, such as a surfer in MauiCaptions and audio repair
A realistic weather event hitting a real city that never happenedCloning your own voice for voiceovers or dubs
Making it look like someone gave advice they never gaveClearly unrealistic scenes, like someone riding a unicorn

Examples from YouTube Help, checked 1 October 2026.

You set it in YouTube Studio when you upload, under Attributes, "AI use". YouTube says disclosing "won't limit a video's audience or impact its eligibility to earn money" (YouTube Help). Channels that keep skipping it risk a label added for them, removed videos or suspension from the Partner Program.

"Our rule at Synthesys is simple: if a viewer could take it for real, we tick Yes," Peter says. "Most of what we post uses realistic AI presenters, so most of our uploads carry the label. That includes digital twins. My twin reading my script is still a scene that was never filmed. Animation, graphics, captions and my own cloned voiceover stay at No, because YouTube doesn't ask for those. The label costs us nothing. Leaving it off can cost a channel."

Which AI rules apply to a faceless channel in 2026?

Four sets of rules matter. YouTube decides what it pays for and what you must label. Google Search judges any blog or site around the channel. The EU now requires deepfakes to be disclosed, and US copyright only protects the parts a human shaped.

RuleWhat it saysWhat it means for a faceless channel
YouTube monetization policiesNo pay for mass-produced template content or AI personas posing as expertsVary every video and keep sensitive advice human
YouTube AI disclosureLabel realistic AI content; the label doesn't cut reach or earningsTick "AI use" for realistic scenes, AI music and realistic AI people
Google Search guidance on AI contentAI content is fine; "many pages without adding value" can be scaled content abuseThe same test applies to your channel's blog or show notes
EU AI Act, Article 50From 2 August 2026, deployers must disclose deepfake image, audio or videoRealistic AI people or events need a clear label for EU viewers
US Copyright Office, Part 2Prompts alone don't make you the author; human-shaped expression can be protectedYour script, edits and choices are what you can own

All pages opened on 1 October 2026. This is a summary, not legal advice.

"AI is already part of every video, from the edit and the captions to the voice," Peter says. "We can't hold it back. It's a matter of time before the licensing questions get sorted."

Not everyone sees it that way yet, and the rules are still arriving. The EU's disclosure duty only started in August 2026. Until things settle, the safe ground is simple. Disclose what looks real, and make sure the ideas and the final cut are yours.

How do you start a faceless YouTube channel with AI, step by step?

Pick a niche you have a view on, list your topics, then script, voice, build, assemble and upload each video. AI does most of the production. You set the angle, check every number in the script and decide what ships. That human check is what YouTube's policy rewards.

The seven steps to start a faceless YouTube channel with AI, with the script check as the human checkpoint

  1. Pick a niche you can add a view to. YouTube wants each video's substance to be "materially varied". A niche where you have opinions makes that easier than one picked for ad rates alone.
  2. List 20 topics before you make anything. If you can't find 20 distinct ones, the niche is too narrow. Every video has to stand on its own.
  3. Write the script, then check every number. AI drafts are fine, and YouTube lists AI script editing as allowed. In both tests below, the script needed fixes before it was ready to post.
  4. Choose the voice. Use a library voice that fits the topic, or clone your own. Cloning your own voice needs no AI disclosure on YouTube.
  5. Build the visuals scene by scene. Ask for a different visual in every scene. A run of near-identical shots is exactly what "templated storylines" looks like.
  6. Assemble it, then watch it with the sound off. Check that captions finish their lines and that key numbers appear on screen, not only in the voice.
  7. Upload it honestly. Use 16:9 for long-form and vertical for Shorts. Verify your account for anything over 15 minutes, and answer the "AI use" question truthfully.

How do faceless channels make money besides ads?

Inside the YouTube Partner Program, you can earn from watch-page ads, Shorts feed ads, memberships and Supers. Outside it, faceless channels use sponsorships, affiliate links and their own products. Paid deals need YouTube's paid promotion box ticked, so viewers see a disclosure label.

Those Partner Program features switch on in YouTube Studio once you're accepted (YouTube Help). Sponsorships and affiliate links don't need the Partner Program at all.

If a video includes "branded content, sponsorships, endorsements, or other commercial relationships", YouTube asks you to select the paid promotion button. That adds a label at the start of the video (YouTube Help).

What does a faceless video look like when an AI video agent makes it?

You type one sentence and get back a script and a timed storyboard you can edit. With "Confirm before running" on, nothing renders until you approve, and the credit cost is on the approve button. The agent then voices the video, builds each scene, adds captions and music, and cuts it.

Synthesys is an AI video agent that replaces traditional video production end to end. It has a Faceless story format, typically 30 seconds to 3 minutes and up to 10 minutes in the app. For longer videos there is a Full YouTube video format of up to 30 minutes, and everything exports in 9:16, 16:9 or 1:1 (Synthesys).

JobWith separate toolsWith one AI video agent
ScriptA writing assistantWritten from your brief, editable
Scene planA doc or spreadsheetA storyboard of named, timed scenes
VoiceA text-to-speech or cloning tool400+ voices in 140+ languages, or your clone
VisualsA stock library or video modelEach scene routed to a model
CaptionsA captioning toolTimed to the voice
MusicA music libraryScored to the edit
EditsA video editorScene, line or voice changes by chat
UploadYouTube StudioStill YouTube Studio: download, then upload

Agent features from the Synthesys product page.

Our test: brief to finished cut in 6 minutes 32 seconds

Here is the one-line brief we gave the agent:

A 60-second faceless YouTube explainer on how compound interest works, for complete beginners, with one calm narrator and a different visual in every scene.

We used the Faceless Story format and left everything else on its defaults (full test).

StageTime after the briefWhat we saw
Storyboard readyby 0:4012 scenes, 60 seconds planned, 9:16
Scene 10 fixed in chat3:21One wrong number corrected before anything rendered
We approvedabout 5:00The cost was on the Generate button
First cut6:321:01 long, 1080×1920, with narration, captions and music

What we changed. One line. Scene 10 said $100 at 10% a year "can pass seven hundred after twenty years". It reaches $672.75 at year 20 and only passes $700 in year 21. One chat message fixed it, at no charge, because nothing had rendered yet.

What worked. The narration matched the script word for word. We asked for a different visual in every scene and got 12 different shots for 12 scenes.

What looked off. Each caption showed only the first seven or so words of its line. The dollar amounts were spoken but never shown on screen. One shot of a man fanning out bills didn't match "one hundred dollars".

Our second test: a faceless Short about Athens rents

On 1 October 2026 we briefed a faceless Short on a topic people search: why rent in Athens feels impossible.

It starts on the home screen. Pick a template, here Brainrot Edit, and a style, here Storytime. Then describe the video in a sentence or two.

Synthesys home screen: (1) the Brainrot Edit template, (2) the Storytime style, (3) a one-line brief for the Athens Short, typed but not sent

Our real brief was longer. We gave the agent the facts to use, with a source named for several of them. It also set the format, 30 seconds in 9:16, and one firm rule: "No people on camera: nothing human appears".

Synthesys chat showing our Athens brief: (1) brainrot split-screen, Storytime style, faceless, (2) the no-people rule and 30 seconds in 9:16, (3) sent at 12:10 PM

Here is what the app recorded for this run.

StageWhat the app showed
Brief sent12:10 PM
Storyboard17 scenes, each planned at about 2 seconds
Word-limit warnings14 of 17 lines flagged as too long for their scene
Approve stepNone: this run went straight from brief to render
Agent work time4 minutes 58 seconds, brief to finished video
Finished Short51.3 seconds at 1080×1920; scenes ran 1.9 to 4.2 seconds each

We asked for 30 seconds and got 51. The storyboard had warned us: 14 lines held more words than their 2-second scenes fit. Most rendered scenes ran longer than planned, up to 4.2 seconds.

Synthesys editor on scene 15 of 17: (1) the scene planned at about 2 seconds, (2) the 17-scene timeline, (3) the line "So Athens froze new Airbnb registrations", (4) final video ready at 51.3 seconds

This run didn't pause for approval, so it has no cost screen. The composer has two modes, "Confirm before running" and "Auto-run without asking", and this one went straight through. The next screen shows the approve step in another faceless project of ours.

Synthesys approve step in a compound-interest project: (1) the approve card with a 120-second final video, (2) the Generate button with its credit cost, (3) the banner waiting for approval

In that project we asked for 90 seconds, and the approve card said 120. Check the length on the card before you spend anything.

The Athens Short had no chat edits, so we made one on its follow-up. "Why Athens Housing Costs So Much: Part 1" is a 33-scene plan we never rendered. Its scene 15 said foreign buyers made "nearly 40 percent", without saying of what.

We asked the agent to cut the number. It rewrote the line in 4 seconds and rendered nothing. The planned visual for that scene still read "40 percent", so check the visuals after a script edit too.

Synthesys chat edit on the unrendered Part 1: (1) our request to cut an unsourced number, (2) the agent's reply, (3) the new scene 15 line, (4) the planned visual still labelled "40 percent"

Here is the finished Short, in the app and as the file it produced.

Synthesys editor with the finished Athens Short: (1) the agent worked for 4 minutes 58 seconds, (2) final video ready at 51.3 seconds, (3) a caption reading "Petralona jumped fifteen point six", (4) Publish and Download

Four unedited frames from the Athens test Short: two off-topic maps at 0:00 and 0:03, and two well-matched shots for Petralona rents and the 42% price crash

Then we did the human part: we checked every number in the narration against public sources.

What the Short saysWhat we found
Athens rents average €9.8 per m², up 5.4% a yearCorrect, RE/MAX Greece data for 2026 (Proto Thema)
Petralona rents up 15.6%, Kypseli up 14.3%Correct, same source
Apartment prices rose 5.2%Athens figure for Q1 2026. By Q2 2026 it was 5.0% (Bank of Greece)
Prices once crashed 42%, between 2007 and 2017Close. The 42% fall is national, from Q3 2008 to Q3 2017 (Alpha Bank)
"So Athens froze new Airbnb registrations"The freeze is real: central districts 1 to 3, from 1 January 2025 to the end of 2026. The "so" is wrong, and our brief didn't say it. The freeze targets today's housing shortage, not the old crash (Greek City Times)
Search counts ("282 searches") and suburb asking prices (€3,500 and €4,170 per m²)We couldn't find a source for either. Both came from our brief, and the Short drops the brief's "up to" and "per listing". We'd cut these lines before publishing

Three claims were exactly right, two were close but dated or loosely labelled, and two had no source we could find. Every number came from our own brief. The agent kept to it, trimmed a few qualifiers and added one false "so".

The lesson for step 3: check your brief as hard as the script.

What looked off on screen. The opening shot is a map of Central Asia, and "In 2026" plays over a map of North Africa. Neither is Athens. The Petralona and price-crash shots matched the words well.

The brief ruled out people and hands. Hands knead slime in the lower loop for most of the Short.

Up top, a hand writes "China" in a notebook at 0:01. A man at a trading screen appears at 0:16, and someone holds a For Sale sign at 0:25.

Captions spelled numbers out as words, such as "FIFTEEN POINT SIX", rather than showing 15.6%.

After the render, the agent's own top suggestion was to speed up scenes 5 to 8, which it said "feel slow".

"Before I post it, I'd cut the two lines we couldn't source and drop the false 'so'," Peter says. "Then I'd ask for numbers on screen as digits, swap the two wrong maps and hold it to 30 seconds. The agent built the Short. Checking the facts was my job."

What mistakes stall most faceless channels?

Five show up again and again: lookalike videos, reading other people's work, unchecked scripts, the wrong shape and no sign of a person. YouTube's policy targets the first two and the last. Our own tests caught the third.

  • Same structure, new topic. Same intro and outro is fine, YouTube says, if "the bulk of your content is different". The same storyline template every time is not.
  • Reading other people's work. Content that "exclusively features readings of other materials you did not originally create" isn't monetizable. Turning articles into narration needs your own commentary on top.
  • Trusting the first draft. An AI draft can state a wrong number confidently, as ours did, or repeat one from your brief. Read the storyboard for facts, not just for order.
  • Leaving the default shape. Our faceless test plan defaulted to 9:16. Ask for 16:9 in the brief when the video is for the main feed.
  • Leaving no sign that a person made it. A TeamYouTube FAQ on reused content, viewed 333,000 times, says reviewers check "how you created, participated in, or produced your content" (YouTube Help Community). Using AI doesn't fail that check on its own, the FAQ says. Yet a volunteer Gold Product Expert there wrote in March 2026 that the denials he sees are "either for copyright or AI" (YouTube Help Community).

What are the limits of a faceless channel built with AI?

AI can produce the video, but it can't supply the angle YouTube pays for. Expect to fix captions, on-screen numbers or loose footage on early cuts. Some topics, like AI personas giving financial or health advice, won't be monetized however good the video is.

  • First cuts need a human pass. In our compound-interest test, captions showed only the first words of each line, and some footage didn't match the narration.
  • A brief isn't a guarantee. Our Athens brief asked for 30 seconds and no people. The Short ran 51 seconds and showed hands for most of its length.
  • Format lengths differ. The Faceless story template runs up to 10 minutes, and the public page lists 30 seconds to 3 minutes as typical. For 10 to 30 minutes, use the Full YouTube video format.
  • YouTube upload stays manual. On the Agency plan and up, Synthesys publishes to TikTok, Instagram and Facebook from inside the app (Synthesys). For YouTube, you download the video and upload it in YouTube Studio.
  • Sensitive topics. An AI narrator presenting as a doctor, lawyer or financial adviser rules out monetization under YouTube's AI persona rule.

"We shipped AI voices in 2020, avatars in 2021 and our own lip-sync model in 2023," Peter says. "Each time, a better model came along. What never went out of date was the step where a person reads the plan and decides. That's why the agent shows you the storyboard before it spends anything."

Try it

Synthesys turns one sentence into a script and storyboard you can edit, then into a finished video with voice, captions and music. Faceless stories and full YouTube videos are both built-in formats, in 16:9, 9:16 or 1:1. Start a faceless video in Synthesys.

On the home screen, Faceless Story sits under Marketing & Social, and Brainrot Edit under Trending.

Read next: What is an AI video agent?, how to make long AI videos without stitching clips and why AI video credits get wasted. See every format on the Synthesys Video Agent page.

Sources

All pages opened and checked on 1 October 2026.

Frequently asked questions

Can faceless YouTube channels make money in 2026?
Yes. A faceless channel can join the YouTube Partner Program with 1,000 subscribers. It also needs 4,000 public watch hours in 12 months, or 10 million Shorts views in 90 days. Each video still has to add original value, or YouTube treats it as inauthentic content.
Does YouTube demonetize AI-generated videos?
Not for using AI on its own. YouTube won't monetize AI content made with generic or unoriginal templates that looks mass-produced. It also won't monetize AI personas posing as experts on health, legal, financial or political topics.
Is YouTube automation allowed?
Yes. YouTube's policy says that if you use automated tools or templates, the final video must still show your creative vision. It also has to give viewers educational or entertainment value.
Do I need to disclose an AI voice on YouTube?
Not for cloning your own voice for voiceovers, which YouTube lists as needing no disclosure. Tick "AI use" when AI makes a real person say something they didn't, or shows a realistic scene that never happened.
Can I upload faceless videos longer than 15 minutes?
Yes, once your Google account is verified. Unverified accounts can upload videos up to 15 minutes. Verified accounts can upload up to 12 hours or 256 GB, whichever is less.
How many views does a faceless channel need to get monetized?
On the long-form route you need 4,000 watch hours, which is 240,000 minutes in 12 months. At 4 minutes watched per view, that's about 60,000 views, plus 1,000 subscribers.

Keep reading

All posts →