AI video glossary: 21 faceless YouTube and AI video terms explained
By Arsène Huot, founder of TubeTube · Published · Updated
21 AI video and faceless YouTube terms, defined
Faceless YouTube channel
A channel that publishes videos with no on-camera presenter. The content is narration or music over visuals (animation, generated scenes, stock or b-roll), which makes it easy to produce at scale. How to make faceless videos.
YouTube automation
Running a YouTube channel where the production is handled by tools or freelancers instead of the owner on camera: script, voice, visuals and edit. With AI, one pipeline can produce the whole video, while the owner still picks the topics, checks each upload and publishes it. YouTube's monetization rules apply as usual: mass-produced, generic or repetitive uploads are not eligible for ad revenue (see inauthentic content below). AI YouTube automation.
Text-to-video
Turning written text (a script, a story or lyrics) into a finished video automatically: the text drives the voice or song, the storyboard, the generated scenes and the final edit. Story to video.
RPM (revenue per mille)
What a creator actually keeps per 1,000 video views, after ad fill rate, non-monetized views and YouTube's revenue share. RPM is almost always far lower than CPM and is the number that decides real income. RPM by country. Source: YouTube Help: Understand ad revenue analytics.
CPM (cost per mille)
What advertisers pay per 1,000 ad impressions. It's the big headline number, but it's not what the creator keeps. The gap between CPM and RPM comes from fill rate, non-monetized views and the platform's cut. CPM vs RPM. Source: YouTube Help: Understand ad revenue analytics.
Character drift (identity drift)
When a generated character's face, clothing, proportions or art style change from one scene to the next, so a multi-scene video looks like it stars several different people. It's the most common giveaway of low-effort AI video. How to fix character drift.
Storyboard
The plan that breaks a script or song into a sequence of scenes, each a few seconds long, so the visuals can be generated and timed to match what's being said or sung at that moment.
Scene
One short shot in an AI video, typically a few seconds long, made of a generated image that is then animated. A multi-minute video is built from many scenes stitched together.
Reference image
An image fed to the generator so it keeps a character, world or prop consistent across scenes. Pinning a few reference images is one of the strongest ways to prevent character drift. Consistency methods.
Video engine
The AI model that animates a still image into motion, such as Kling, Google Veo or Hailuo. Engines differ in resolution, scene length, cost and sound: some render silent clips, and three can speak lines on camera in dialogue mode. Kling vs Veo vs Hailuo.
Dialogue mode
A TubeTube mode where named characters say their lines on camera: you write the dialogue into the script and a speaking video engine renders the speech inside the clip, lips moving, instead of a narrator reading over the picture. Three engines can speak lines on camera in TubeTube's dialogue mode: Veo 3.1 (12 script languages), Kling 3 (English, Chinese, Spanish) and Gemini Omni Flash (English supported by Google, French verified by TubeTube). Hailuo 2.3, Seedance 2 and Kling 2.5/2.6 render silent clips. Outside dialogue mode, TubeTube removes the engines' own audio from every clip, including Veo's and Omni Flash's, and the pipeline mixes ElevenLabs narration or Suno music over them. Inside dialogue mode, only the shots where a character speaks keep the engine's sound; narrator lines are voiced by ElevenLabs over a clip without it. Gemini Omni Flash in dialogue mode.
Upscale
Raising a finished video's resolution, for example to 2K or 4K, with an AI upscaler. TubeTube's pipeline has an upscale step, but it is switched off: videos come out at the video engine's own resolution, 720p to 1080p, and there is no 2K or 4K export.
TTS (text-to-speech)
AI narration: a voice model reads a script aloud. Modern TTS (for example ElevenLabs) offers many voices filterable by language, gender and age, with audio previews.
Dubbing
Re-voicing a finished video into another language while keeping its tone and timing. Dubbing one video into several languages lets a single piece of content reach several ad markets at once.
Lyric video / music video
A video built around a song rather than narration. The audio is analysed so scene changes land on the song's phrasing, and the visuals are generated to match the mood of the track. AI music video generator.
Suno
An AI music generator that turns a short description into a finished song with vocals, or sings lyrics you write yourself. TubeTube uses Suno to compose and sing the lyrics of a music video (27 credits a song) and to make a sung cover of a track you upload; a song you already made on Suno can also become a video. Suno song to video. Source: Suno Help: Can I use my own lyrics?.
Watch time
The total minutes people spend watching a channel's videos. YouTube weighs watch time heavily for reach and monetization, often more than raw view count.
Made for kids (COPPA)
The audience setting YouTube requires on every channel and video. Content counts as made for kids when children are its main audience, or when its characters, songs, stories or activities show an intent to target children. YouTube asks for it to comply with the US Children's Online Privacy Protection Act (COPPA) and other laws, and says a wrong setting can have consequences on YouTube and under COPPA. Made-for-kids videos get no personalized ads and lose features such as comments, the notification bell and end screens, so they usually earn a lower RPM. Sources: YouTube Help: Determining if your content is made for kids; YouTube Help: Set your channel or video's audience.
Inauthentic content (YouTube policy)
The name YouTube gave on 15 July 2025 to its former 'repetitious content' monetization policy: mass-produced, generic or repetitive uploads are not eligible for the YouTube Partner Program. YouTube's own policy page lists using AI to visualize a character and story you invented as acceptable, and excludes AI content made from generic templates that gives an impression of mass production. Source: YouTube channel monetization policies.
Altered or synthetic content disclosure
YouTube's label for realistic content made or changed with AI. Creators must turn it on when a video makes a real person appear to say or do something they didn't, alters footage of a real event or place, or shows a realistic scene that never happened. Clearly unrealistic content, such as a fully animated video, and production help, such as an AI-written script or thumbnail, don't require it. YouTube says the label doesn't limit a video's audience or its eligibility to earn money, and it may apply an AI label itself when its systems detect AI-generated content. Source: YouTube Help: Disclosing use of GenAI content.
Credits
TubeTube's usage-based currency, worth about $0.011 each at the base rate. A video reserves an estimated number of credits up front and the unused part of the hold is refunded, so you pay for what is actually produced. A 2-3 minute video costs a median 807 credits, about $9 at the base rate of $0.011 per credit; half of them cost between 685 and 1,129 credits ($7.50-12.50), measured on 246 videos in August-September 2026. The exact quote is shown before launch and unused credits are refunded.
Once the vocabulary makes sense, see what the pipeline actually produces across formats on the use cases page, or read what TubeTube is and who builds it.
Frequently asked questions
What is the difference between RPM and CPM on YouTube?
CPM is what advertisers pay per 1,000 ad impressions, before YouTube takes its share. RPM is what the creator actually earns per 1,000 video views, after YouTube's share and across all revenue sources, so it is almost always far lower than CPM and is the number that decides real income. Both definitions come from YouTube's help article on ad revenue analytics.
What is character drift in AI video, and how do you stop it?
Character drift is when a generated character's face, clothing or art style changes from one scene to the next. It comes from generating each scene on its own. It is reduced by locking a reference image, keeping the character description word for word identical, holding one visual style, and generating scenes in order with the earlier images fed back in as context.
What is dialogue mode in AI video?
It is the mode where characters speak their lines on camera, with the speech rendered inside the clip by a speaking video engine rather than narrated over it. Three engines can speak lines on camera in TubeTube's dialogue mode: Veo 3.1 (12 script languages), Kling 3 (English, Chinese, Spanish) and Gemini Omni Flash (English supported by Google, French verified by TubeTube). Hailuo 2.3, Seedance 2 and Kling 2.5/2.6 render silent clips. Outside dialogue mode, TubeTube removes the engines' own audio from every clip, including Veo's and Omni Flash's, and the pipeline mixes ElevenLabs narration or Suno music over them.
Can TubeTube upscale videos to 2K or 4K?
No. The pipeline has an upscale step, but it is switched off, so every video comes out at the video engine's own resolution, 720p to 1080p. There is no 2K or 4K export.
Do AI videos need a disclosure label on YouTube?
Only when they look realistic. YouTube asks creators to disclose AI-made or altered content that makes a real person appear to say or do something they didn't, alters footage of a real event or place, or shows a realistic scene that never happened. Clearly unrealistic content, such as a fully animated video, and production help such as an AI-written script don't require it, and YouTube says the label doesn't affect a video's reach or its eligibility to earn money. YouTube may also add an AI label itself when its systems detect AI-generated content.
Is YouTube automation with AI allowed?
Yes, YouTube allows AI-made videos, but its monetization policy on inauthentic content excludes AI content made with generic or unoriginal templates that gives the impression of mass production. The same policy lists using AI to visualize a unique character and narrative you invented as acceptable. An original channel qualifies; a stream of near-identical uploads does not.