Short answer: most tutorials assume you already know a dozen words: reference sheet, LoRA, seed, image-to-video, lip-sync, drift, AI label. This page explains the 30 you will actually meet, grouped in the order you meet them: the character, the pictures, the video, posting and the rules, and money. Each one is a sentence or two in plain English, with a link to the full guide and, where we have one, a number from our own production logs.
The numbers come from Influgen's own logs, published at /data and re-read on 7 November 2026: 672 frame reviews of 124 posts by 10 AI influencers (18 to 27 October 2026) and our logged provider spend, plus the first full November week (198 frame reviews, 2 to 8 November 2026, re-read on 10 November 2026). They describe one product's pipeline, not AI images in general. Rule terms are explained in general terms and link to guides that quote each platform's own page; nothing here is legal advice.
Thirty words most AI influencer tutorials assume you already know.
At a glance
Group
Terms
The character
AI influencer, virtual influencer, faceless channel, persona sheet, reference sheet, face consistency, signature look
shot list, image-to-video, reference-to-video, lip-sync, voice, cost per second
Posting and the rules
AI label, EU AI Act Article 50, ad disclosure, content calendar, reach vs. followers
Money
brand deal, UGC ad, affiliate link, creator rewards
The character
1. AI influencer. A social media account whose face is a generated character, not a real person, run like a creator's account: one personality, one niche, regular posts. Who runs it and how much is automated varies a lot from tool to tool. Start here: how to create an AI influencer.
2. Virtual influencer. The older word, from the era of 3D characters made by studios. Today it is mostly used for stylised or cartoon-like characters, and "AI influencer" for photo-real ones made with image models. The full difference: virtual influencers vs AI influencers.
3. Faceless channel. An account with no face at all: voice-over, b-roll, text, hands, a product. It is the other common route for people who don't want to be on camera. Which one to start: faceless content vs an AI influencer.
4. Persona sheet. One page that says who the character is before you draw anything: age, city, job, what she likes, how she talks, what she would never post. It is what keeps posts sounding like one person. A fill-in template: the persona sheet.
5. Reference sheet (reference images). A small set of approved pictures of the character, usually the face from the front and side and the full body, that every new picture is drawn from. Influgen draws every frame from each influencer's reference images. How to write prompts around them: AI influencer prompts.
6. Face consistency. Whether the character still looks like the same person from post to post. It is the problem everyone worries about first, but with good references it was not our biggest one: "not the character" was the main reason for only 3% of the frames our critic turned down. Still, check by number, not only by eye: in 2 to 8 November 2026 a separate face check turned down 14 frames the critic had scored 8 or 9, all measuring 0.34 to 0.45 against her own photos, at or just under her line. More: why face consistency matters.
7. Signature look. The two or three things that make the character recognisable at a glance even when the face is small: a hair colour and cut, a jacket, a piece of jewellery. Keep it the same across a post.
The pictures
8. Prompt. The text you give an image or video model to say what to make. Short, concrete prompts that describe the scene and the action usually beat long lists of style words. Templates: AI influencer prompts.
9. Text-to-image. Making a picture from a prompt alone. A picture from GPT Image 2.5 cost us $0.021 on average (1,003 calls, 2 October to 1 November 2026), about a fifth of what one second of video cost.
10. Image editing (image-to-image). Giving the model a picture plus an instruction ("same woman, now in a café") instead of starting from nothing. This is how most tools keep a character: the reference images go in with every request.
11. LoRA. A small add-on file trained on many pictures of one face or style, which an open image model loads to draw that face again. It is one way to get consistency, mostly in self-hosted setups. You don't need one if your tool draws from reference images; Influgen doesn't train LoRAs.
12. Seed. The random number a model starts from. The same prompt with the same seed gives a similar picture; a new seed gives a new variation. Useful for trying alternatives, not a way to lock a face.
13. Frame. One still picture in a post: one photo in a carousel, or the start picture a video clip is made from. Our figures count frames, because each one is judged on its own.
14. The "AI look". Plastic skin, a cinematic colour grade, everything a bit too perfect. It was the main reason for 18% of our rejections, the second biggest after "wrong scene". Ten fixes: how to make AI influencer photos look real.
Most of the words describe one of five stages: character, pictures, video, posting, money.
15. Drift. When the outfit, hair, props or place change between pictures that are supposed to be the same moment. "Doesn't match the other frames" was the main reason for 9% of our rejections. The fix is the same reference sheet and the same look written into every shot.
16. Critic (review pass). A check of each picture before it is posted: does it show what the shot needs, does it look like the character, does it look real. Influgen's critic is an AI vision model; it kept 323 of 672 frames (48%) in our first cut, and the most common reason for a reject was a fine picture of the wrong thing (38%), not extra fingers (6%). The full numbers: we reviewed 672 AI influencer frames. In the first full November week (2 to 8 November 2026) it kept 136 of 198 (69%), and the most common reason for a reject became a frame that shows the payoff too early or not at all (22 of 62 notes): what our critic rejected, week 2.
The video
17. Shot list (storyboard). The plan for a video, shot by shot: what is in the frame, what moves, how long. Writing it first is cheaper than fixing clips afterwards. How: AI video prompts, shot lists and camera moves.
18. Image-to-video. Making a short clip that starts from a still picture, so the character and the place are already right. What we paid per second of finished clip over 30 days: Kling O3 Standard $0.14, Kling v4 $0.157.
19. Reference-to-video. Making a clip from reference pictures of the character instead of one exact start frame; the model places her in the scene itself. Wan 3.0 reference-to-video cost us $0.10 a second over the same 30 days. A third route, motion control, copies the moves of a real clip onto the character; it is how trend dances are made, and Kling 2.6 Pro cost us $0.112 a second in our 7 to 8 November 2026 test: how to make an AI influencer dance video.
20. Lip-sync (talking head). A clip where the character's mouth moves to a voice track, so she appears to speak. Kling AI Avatar v2 Pro cost us $0.115 a second. How talking videos are made: how to make a talking AI influencer video.
21. Voice. The synthetic voice the character speaks with. Use a voice designed for the character or a stock voice; never copy a real person's voice without their written consent. Who owns what: copyright, likeness and your rights.
22. Cost per second. The honest way to compare video models, because clips differ in length and you pay for retries too. Pictures are an order of magnitude cheaper, which is why it pays to get the still right before you animate it. Work out a whole reel with the video cost calculator.
Posting and the rules
23. AI label. The platform's own marker that a post (or a whole profile) is AI-generated. Instagram, TikTok and YouTube each have one, and each wants realistic AI people labelled. Where each switch is: how to add an AI label; what Instagram does with AI profiles: the AI-generated profile label.
24. EU AI Act Article 50. The EU rule that, from 2 August 2026, makes AI-generated or manipulated images, audio and video that look real ("deepfakes") be disclosed as such. It matters if your audience is in the EU. The AI disclosure checker gives you the line to add for each platform.
25. Ad disclosure (#ad). A separate rule from the AI label: if a post is paid, gifted or has an affiliate link, it has to say so clearly, AI or not. In the US this is the FTC's endorsement rules. Both rules together: do AI influencers need to disclose?.
26. Content calendar. The plan of what goes out when, usually a week or a month ahead, so the account posts steadily without a scramble every day. Templates: AI influencer content calendar.
27. Reach vs. followers. Reach is how many accounts saw a post; followers are the people who chose to see the next one. They are not the same thing: one of our reels reached 3,320 accounts and got 5,263 views, and 0 follows. Ten mistakes like this: AI influencer mistakes beginners make.
Look at every frame at full size before it goes out.
Money
28. Brand deal. A brand pays the account to feature its product. Brands want an audience first, and the post must be marked as an ad. An honest guide: can an AI influencer get brand deals?.
29. UGC ad. A short, creator-style video ad ("user-generated content") that a brand runs from its own account. Unlike a brand deal, it doesn't need your followers. The difference: AI UGC vs AI influencers; what one costs: an itemized AI UGC ad.
30. Affiliate link and creator rewards. An affiliate link pays a commission when someone buys through it; creator rewards (TikTok Creator Rewards, YouTube's Partner Program and similar) pay for views once an account meets the platform's thresholds. Both have rules for AI content. Guides: AI influencer affiliate marketing, how many followers you need to make money.
Quick answers
What is the most important term to learn first? Reference sheet. Almost every consistency problem (face, look, drift) comes back to whether every picture is drawn from the same approved references.
Do I need a LoRA to make an AI influencer? No. A LoRA is one way to keep a face, mostly with open models you run yourself. Tools that draw from reference images, like Influgen, don't need one.
What's the difference between an AI label and an #ad? The AI label says the content is AI-made; #ad says someone paid or gave you something. A sponsored post of an AI influencer needs both.
What does image-to-video cost? We paid $0.10 to $0.157 per second of finished clip over 30 days, depending on the model, against about $0.02 for a picture. A 20-second reel is several clips plus retries.
Is "virtual influencer" the same as "AI influencer"? Mostly, in everyday use. "Virtual" leans to stylised 3D characters made by studios, "AI" to photo-real characters made with image models.
Make your own
You don't need to master all 30 words before you start. Influgen turns them into a team: you describe the person, and her agents write the persona sheet, keep the reference sheet, plan the shots, check every frame with a critic and put the AI label on every post. Create your own influencer: free to create, no card, a 7-day trial with 50 credits.
Influgen gives your influencer a team that plans, creates and posts on its own. Create your own influencer free. The first 1,000 accounts get 100 extra credits.
Setting up a character took about 5 minutes, one picture 24 seconds and a 12-second video about 5 minutes. Our own timings for 1,350 pictures and 42 videos, and the two steps that take far longer than the making.
The ten mistakes that cost new AI influencer accounts the most, from our own review logs: half of all frames turned down, 38% for showing the wrong thing, a reel with 5,263 views and 0 follows, and TikTok's one-minute rule.
Twelve things to settle before an AI influencer posts for the first time: one niche, an original face, a name and handle, a reference sheet, an honest bio, the AI labels on Instagram, TikTok and YouTube, a budget, two weeks of posts, a review step and a link in bio. Each step links a free tool or a guide.
