An AI influencer is a fictional person who appears again and again, so the work is less about one good image and more about keeping one identity across hundreds of posts. On an 8frame board that means one place for everything fixed: a face still, a character sheet, a wardrobe set, a few settings, one voice, and video nodes that always reference those stills, such as Happy Horse with 1 to 9 character references (95 credits for 5 seconds at 720p). The limit people do not expect: 8frame has no voice cloning, so the "voice" is a choice you write down and reuse (one TTS voice, or one consenting human's recordings), and it drifts if you do not.
For the ways brands use AI creators alongside human ones, see AI for influencer marketing. For the disclosure rules specific to synthetic people, see AI influencer disclosure rules. This page is the board.
TL;DR
| Step | Node or model on 8frame | Fixed price (credits) | Input it needs |
|---|---|---|---|
| 1. Identity notes | sticky notes on the board | none | your brief |
| 2. Face | Nano Banana Pro or GPT Image 2 | 21 at 1K or 2K; 29 square, 20 other (GPT Image 2 high) | description |
| 3. Character sheet | ShotGrid, or Nano Banana Pro per view | 42 per 3x3 sheet at 4K; 21 per view | the face still |
| 4. Wardrobe | Nano Banana Pro with references | 21 | sheet + clothing references |
| 5. Settings | Nano Banana 2 or Seedream 5 Lite | 11 at 1K; 6 | the character + location prompt |
| 6. Voice | ElevenLabs TTS | 13 per started 1,000 characters | script |
| 7a. Clips from references | Happy Horse, reference-to-video | 95 for 5 s at 720p, 189 at 1080p | 1 to 9 character references |
| 7b. Talking clips | Pruna Avatar | 3.375 per second at 720p | portrait + audio |
| 7c. Your movement, their body | Kling 2.6 Pro Motion Control | 18.9 per second of the driving clip (5 s = 95) | your own clip + a character image |
The board, node by node
There is no AI influencer template live on 8frame yet. The live templates that cover parts of it are CHARACTER POSES and CHANGE OF CLOTHES. Build the rest on one board, so every future post starts from the same references.
Step 1: identity notes
Before any image, write the fixed facts on sticky notes in one frame on the board: name, age range, look, the three outfits they rotate, where they live, how they talk, what they never do. These notes are what you check every new image against. A recurring character drifts one small change at a time; the notes are how you notice.
Keep the person fictional. Do not base them on a real person, a creator you admire or a celebrity. The FTC's reviews rule Q&A says using a celebrity's avatar to praise a product without permission violates the rule if consumers would believe the celebrity endorsed it, and lookalikes raise the same question.
Step 2: the face
Make the defining portrait on Nano Banana Pro (21 credits at 1K or 2K) or GPT Image 2 (high quality, the default: 29 credits square, 20 for other ratios). Front-facing, neutral expression, plain background, even light, shoulders visible. Choose it the way you would cast a person: you will see this face every week.
Front-facing portrait of a fictional woman in her early thirties, freckles, shoulder-length auburn hair with a centre parting, small gold hoop earrings, neutral relaxed expression, plain light grey background, soft even light, shoulders visible, photographic, no text
Step 3: the character sheet
Turn the face into views: front, three-quarter, side, back, full body, three expressions. ShotGrid does nine views in one 4K image for 42 credits, with 1 to 10 references; Nano Banana Pro does one view per node at 21. Separate views are better references for video, because each reference slot then holds one clear view. The details are in character sheet to video.
Step 4: wardrobe
Two or three outfits that repeat, like a real person's. Generate each outfit on the character with Nano Banana Pro, connecting the sheet and a clothing reference (a flat-lay photo of a real garment you own or sell, or a generated one). CHANGE OF CLOTHES is built for this step.
Lock each outfit as a reference still. Next week's post uses the still, not a new description.
Step 5: settings
Three or four places the character keeps returning to: a kitchen, a desk, a gym, a street. Make each setting with the character in it on Nano Banana 2 (11 at 1K) or Seedream 5 Lite (6, up to 14 references). Recurring places do as much for recognition as the face.
Step 6: the voice
8frame has no voice cloning. Two ways to keep one voice:
- One TTS voice. Pick one ElevenLabs voice and write its name and settings on the identity notes. TTS is 13 credits per started 1,000 characters, up to 5,000 per generation.
- One human voice. A real person who records the scripts and consents in writing to the use, for as long as the character runs.
Step 7: clips
Every clip references the sheet, the outfit and the setting. Never chain from last week's clip.
- Happy Horse, reference-to-video with 1 to 9 character references: 95 credits for 5 seconds at 720p, 189 at 1080p. The node opens at 1080p, 16:9; set 9:16 for Reels and TikTok.
- Talking to camera from a portrait and the voice: Pruna Avatar (3.375 per second at 720p, up to 35 seconds), OmniHuman v2, the default OmniHuman (14.175 per second of audio, 71 for 5 seconds, 4 to 35 seconds), or Hailuo H3 Max Lip Sync (54 per 5 seconds at 768p, audio clipped at about 14.8 seconds). See best AI talking avatar generator.
- Your movement, their body. Film yourself (body or upper body with the head fully visible, camera fixed), then run Kling 2.6 Pro Motion Control with the character image: 18.9 credits per second of the driving clip, 3 to 30 seconds (10 in image-orientation mode). It moves the character with your motion; it does not make the character speak, so the voice goes on the Video Editor sound layer. Steps in motion control from your own video.
For the lip-sync and motion-control models we quote price and settings, not results.
Join clips on the Video Editor timeline. The Montage node outputs no audio track, so the voice would be lost.
Which model for which step
| Job | Model | Price | Key limit |
|---|---|---|---|
| Face | Nano Banana Pro | 21 (1K, 2K), 42 (4K) | up to 10 references |
| Sheet in one image | ShotGrid | 42 at 4K | nine views in one reference slot |
| Outfit and setting stills | Nano Banana Pro, Nano Banana 2, Seedream 5 Lite | 21, 11, 6 | Seedream 5 Lite takes up to 14 references |
| Moving clips | Happy Horse | 95 (720p), 189 (1080p) for 5 s | 1 to 9 character references |
| Moving clips, cheaper | Grok Imagine reference | 35 (480p), 48 (720p) for 5 s | 1 to 10 images; audio always on |
| Talking | Pruna Avatar | 3.375 per second at 720p | up to 35 s; keeps the portrait's shape |
| Your own movement | Kling 2.6 Pro Motion Control | 18.9 per second of the driving clip | 3 to 30 s; voice added on the timeline |
Where it breaks
- Slow drift. Each new image copies the last one's small errors. Reference the sheet every time and compare with the identity notes.
- Voice drift. A different TTS voice, speed or language setting changes the person. Write the settings down.
- A face that resembles someone real. Check it before you build a following on it.
- Outfits described instead of referenced. A text description gives a slightly different jacket every week. Reference the outfit still.
- Close-up faces in video. If a tight face shot fails, try a waist-up frame. Most failures show "Sensitive content detected. Please try changing your prompt or image."; failed runs are refunded.
- Montage. No audio. Use the Video Editor timeline.
Why brands are building these
According to TrendWatching's summary of a Fiverr release (May 23, 2024), searches on Fiverr for "AI influencer" rose 6,305% over six months, and a Fiverr release of December 10, 2025 listed "AI influencer (+43%)" among growing searches. These are searches on a freelance marketplace, not audience results.
FAQ
How do I create an AI influencer that looks the same in every post? Make one face, a character sheet and fixed outfit and setting stills, keep them on one board, and reference them in every new image and clip.
Can 8frame clone a voice for my AI influencer? No. 8frame has no voice cloning. Use one TTS voice with fixed settings, or recordings from one consenting person.
Which model keeps the character in video? Happy Horse takes 1 to 9 character references (95 credits for 5 seconds at 720p). Grok Imagine takes 1 to 10 references (48 at 720p). Neither had a quality test from us; compare on your own references.
Can I base an AI influencer on a real person? Only with that person's written consent for the use, and never a public figure.
Do I have to disclose that the influencer is AI? Platforms and some laws require it in many cases. See the disclosure rules page linked above.
Sources
- 8frame prices, defaults and limits: the 8frame app, checked 2026-11-09.
- FTC, Consumer Reviews and Testimonials Rule: Questions and Answers: https://www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers, checked 2026-11-09.
- TrendWatching, "On Fiverr, searches for 'AI influencer' have skyrocketed by 6,305% in the last 6 months" (2024-05-23): https://trendwatching.com/trends-and-insights/on-fiverr-searches-for-ai-influencer-have-skyrocketed-by-6305-percent-in-the-last-6-months, checked 2026-11-09.
- Fiverr release "Demand for AI Video Creators Surges 66%..." (2025-12-10), via Stock Titan: https://www.stocktitan.net/news/FVRR/demand-for-ai-video-creators-surges-66-as-businesses-race-to-scale-x7syywe5y5ax.html, checked 2026-11-09.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.