Skip to main content
/ai-influencers · 2026 edition

Build a repeatable virtual-brand
pipeline.

The complete pipeline for building virtual brands at scale: identity engineering, ComfyUI production, IP governance, and the distribution flywheel.

Get the free Starter KitBuilt by Anyro · founder of IImagined (est. 2024) · Melbourne

Read the first lesson before you pay.

This is the opening of “Flux & SDXL Setup”, the first lesson of the course. Make a free account and you can read all of it.

Module 1 · Lesson 1

Flux & SDXL Setup

Pick Your Base Model First (October 2026)

Every later lesson depends on one decision: the image model you build your character on. A LoRA only works on the model family it was trained for, and the model's licence decides whether you may use it on a paid account. Choose before you train anything.

One decision every later lesson depends on
  1. 01
    Choose the base model

    Before you train anything

  2. 02
    Check its licence

    It decides whether you may use the model on a paid account

  3. 03
    Build on that family

    A LoRA only works on the model family it was trained for

ModelLicencePaid-account useVRAM (vendor or Comfy figure)Strength
FLUX.2 [klein] 4BApache 2.0YesAbout 8-13 GBFast photoreal images and multi-reference editing
Z-Image Turbo (6B)Apache 2.0YesFits in 16 GBFast photoreal drafts in 8 steps
Qwen-Image / Qwen-Image-Edit (20B)Apache 2.0YesAbout 24 GB in fp8Instruction editing, text in images
SDXL 1.0 and fine-tunesCreativeML Open RAIL++-M; fine-tunes set their own termsYes, within its use restrictions8 GBLargest LoRA and ControlNet library
FLUX.1 [dev] (12B)FLUX.1 [dev] Non-CommercialUnclear: the model is non-commercial, its outputs may be used commercially; read the licenceAbout 23 GB fileLarge existing LoRA ecosystem
FLUX.2 [dev] (32B)FLUX Non-CommercialUnclear: the model is non-commercial, its outputs may be used commercially; read the licenceH100-class unless quantizedTop open-weight quality, for testing

Sources: model cards for FLUX.2 [klein] 4B, Z-Image-Turbo, SDXL and the FLUX.1 [dev] licence.

Which base model can carry a paid account
FLUX.1 [dev] and FLUX.2 [dev]
  • FLUX Non-Commercial licences
  • Paid-account use unclear: the model is non-commercial, its outputs may be used commercially
  • Read the licence before you rely on it
Apache 2.0 models or SDXL
  • FLUX.2 [klein] 4B, Z-Image Turbo, Qwen-Image: Apache 2.0, paid use yes
  • SDXL 1.0: CreativeML Open RAIL++-M, yes within its use restrictions
  • VRAM: about 8-13 GB for FLUX.2 [klein] 4B, 8 GB for SDXL

Source: Model cards and the FLUX.1 [dev] licence

Read the full lesson free

Straight from the free community.

Anyro's latest posts in the free IImagined community: what's working on Instagram, AI tool updates, product and SaaS notes, and weekly challenges. The free Creator Starter Kit comes with a free account, and its dashboard has your invite to the Discord community.

Get the free Starter Kit

Latest posts from the feed, refreshed daily.

  1. Anyro·

    Challenge: the five-shot consistency test

    mini challenge for this week. it takes about an hour and it tells you whether your character is actually consistent. generate your character in these 5 shots: 1. neutral front…

    1 likes
  2. Anyro·

    What's stopping your growth right now?

    quick poll so i know what to write about next. where are you actually stuck on instagram right now? if you pick views, tell me your typical skip rate if you know it. if you pick…

    1 likes
  3. Anyro·

    Poll: where will your character earn first

    poll for anyone building an AI character. where do you expect the first money to come from? I'm asking because it changes what you should build first: fan platform: you need a…

    1 likes
Get the free Starter Kit
01 · The current state

Where it hurts right now.

⏵
Camera fatigue
You're tired of being the brand. Every post is your face, your voice, your time.
☷
Unclear business model
Seeing polished virtual creators without knowing the production, distribution, disclosure, cost, or monetization work behind them.
◌
Broken AI workflows
You've tried ComfyUI. Faces drift, hands break, your "brand" looks different in every render.
$
No path to revenue
No clue how AI accounts monetize without traditional brand deals or a real person to pitch.
02 · Why other paths fail

The graveyard of attempts.

Compare these approaches with the structured curriculum and decide what fits your goals.

midjourney solo
“A few prompts, post the image.”
No consistency, no identity, no production pipeline. Your "brand" is a Pinterest board that the algorithm treats as random images. No watch time, no follow signal.
character.ai persona
“A chatbot you talk to.”
Solves the wrong layer. You don't need a chatbot. You need a content factory. Personality without distribution is therapy, not a business.
paid prompt packs
“Buy prompts without a production workflow.”
Prompts are only one input. Consistency also depends on the identity spec, model, controls, assets, evaluation, motion, and distribution process.
agency outsource
“Pay an agency to be your creative team.”
You're back to trading dollars for hours, just someone else's hours. No leverage, no IP ownership, no scaling. You quit when you stop paying.
03 · What's different here

The line between tactic and system.

Treat AI influencers as a prompt-engineering trick.
Treats your virtual brand as a software product with a spec, build pipeline, deploy step, and revenue model.
Product mindset
Show a single ComfyUI workflow and call it a course.
Full pipeline: identity spec → ControlNet face consistency → motion → multi-platform deploy → governance.
7-step pipeline
Ignore the legal/IP/disclosure mess.
Likeness governance, brand-deal disclosure that survives scrutiny, IP ownership doc templates.
Governance covered
Stop at the post, no distribution plan.
Multi-platform orchestration: Reels, TikTok, Threads, YouTube Shorts, all scheduled, tagged and A/B tested.
4 platforms wired
04 · Named mechanism
The Synthetic IP Stack

A four-layer pipeline for repeatable virtual-brand production.

Most AI-influencer attempts die at layer two: bad look-dev, breaking face consistency, no distribution loop. The Synthetic IP Stack treats your virtual brand as a software product: identity is the spec, ComfyUI is the build pipeline, distribution is shipping, monetization is revenue. Each layer hands clean output to the next.

  1. Identity

    Spec the brand: persona doc, voice, niche, look-dev. The thing the model is trained to be.

  2. Production

    Build and document ComfyUI, consistency, motion, and audio workflows for repeatable batch production.

  3. Distribute

    Multi-platform deploy. Reels, TikTok, Threads, YouTube Shorts, all orchestrated, scheduled and tagged.

  4. Monetize

    Subscription pages and pay-per-view on Fanvue, fan messaging, and brand deals that survive disclosure.

Then the loop starts again at step 01: render → ship.

05 · Catalog outline

9 listed modules. 39 listed lessons.

Lessons are added and updated over time.

Image models and setup, hardware choices, prompting basics, and finding models on Civitai and Hugging Face safely.
Lessons
  • Flux & SDXL Setup
  • Hardware Requirements (GPU Guide)
  • Prompt Engineering 101
  • Navigating CivitAI & HuggingFace
Lessons in each module
1. The Tech Stack
4
2. Character Design
3
3. Content Production
5
4. Social Strategy
3
5. Subscriptions and Pay-Per-View on Fanvue
4
6. Next-Gen Video and Motion
4
7. Advanced Video and Livestreaming
5
8. The Chatting Business
6
9. Funnel Architecture
5

Module 1 opens with the free preview lesson.

10 working files, tied to lessons.

The resource vault comes with AI Influencers: 3 spreadsheets and 7 templates and checklists, each used in a specific lesson.

  • Character Bible Template

    character-bible-template.md

    A one-page template for your character's identity, master face description, voice, boundaries, pillars and production settings, plus the AI disclosure lines for every bio; fill it in before you generate a single image.

    Module 2, lesson 1

  • Face Consistency and LoRA Dataset Checklist

    face-consistency-and-lora-dataset.md

    A step-by-step sheet covering base model licences, a reference-image test, dataset coverage and curation, captions, a training run log and a prompt matrix for picking the best checkpoint; work through it in order before and during your first LoRA run.

    Module 2, lesson 2

  • Likeness Consent and Records Kit

    likeness-consent-and-records.md

    A decision table for when a real person needs a release, a release checklist to brief your lawyer, the likeness and deepfake laws in force in October 2026, and a records log to keep for every post.

    Module 3, lesson 4

  • Weekly Production Pipeline

    weekly-production-pipeline.md

    A ten-stage weekly checklist from planning to posting, with the review gate every item must pass, the AI label setting for each platform and what you must never automate; run it once a week.

    Module 6, lesson 3

  • AI Disclosure and Compliance Checklist

    ai-disclosure-compliance-checklist.md

    Setup, per-post and sponsored-post checklists for Instagram, TikTok, YouTube, X, Fanvue, OnlyFans and Fansly, with the FTC rules and a link to every official policy page; re-check it every quarter.

    Module 6, lesson 4

  • Content Calendar

    content-calendar.csv

    A four-week spreadsheet with one row per post, columns for AI labels, sponsorship disclosure and review status, and a filled-in example week; type your start date once and every date updates.

    Module 7, lesson 5

  • Fan Messaging SOP

    fan-messaging-sop.md

    An operator SOP with word-for-word disclosure lines, offer rules, hard stops and crisis replies, a paid test, a weekly QA sheet and the chatbot laws behind each rule; every operator signs it before getting access.

    Module 8, lesson 3

  • Retention Tracker

    retention-tracker.csv

    A monthly spreadsheet that calculates churn, average lifetime, lifetime value, repeat purchase rate and dispute rate from your fan-platform numbers, and flags months above Fanvue's 1.5% dispute threshold.

    Module 8, lesson 6

  • Brand Deal Pitch Kit

    brand-deal-pitch-kit.md

    Three pitch emails that say up front the creator is AI, a media kit outline, what a virtual creator can and can't claim in a sponsored post, and a contract checklist; run the pre-pitch checks first.

    Module 9, lesson 4

  • Tool and Cost Tracker

    tool-cost-tracker.csv

    A budget spreadsheet listing each tool's plan, price, credits and commercial-use terms, with formulas that turn every billing type into a monthly cost and total your actual spend by month and category.

    Module 9, lesson 5

06 · The implementation map

A roadmap, not a promise.

Suggested milestones for planning your implementation. Timing and results vary.

  1. Weeks 1 to 2

    Spec the Brand

    • Identity doc finalized
    • Look-dev brief
    • 12 reference renders approved
  2. Weeks 3 to 4

    Look-Dev Lock

    • Face consistency stable
    • LoRA trained
    • 30 test images shipped
  3. Weeks 5 to 6

    First Content Batch

    • First week of content rendered
    • Voice clone live
    • Audio mastered
  4. Weeks 7 to 8

    Distribution Live

    • Posting across 3 platforms
    • Initial audience signal
    • Tags optimized
  5. Weeks 9 to 10

    Monetize

    • Fanvue page set up
    • Subscription price and pay-per-view set
    • Disclosure copy live
  6. Weeks 11 to 12

    Review the Loop

    • Batch workflow measured
    • Cross-platform process documented
07 · What you need

Prerequisites, honestly.

The actual stack and the actual time. No "you'll need nothing" lies.

◷
8 to 12 hours per week
Render time, edit, scheduling. Most of weeks 1 to 4 is system setup; ongoing ops drops to ~6h after that.
⌘
Compatible local hardware or cloud GPU
Requirements depend on the model, resolution, and workflow. Compare current provider documentation and costs before choosing.
↗
Cloud option
optional
Cloud GPU services can replace local rendering, but availability, privacy, licensing, and usage costs vary.
◯
Optional voice tool
optional
If you use voice generation or cloning, review current pricing, consent requirements, licensing, and platform policies.
△
Capture One or Lightroom
optional
For color-grading consistency. Free DaVinci Resolve also works fine.
∞
Patience: the first 4 weeks are setup
Identity + look-dev is a one-time investment. Months 2+ are pure shipping. Don't quit at week 3.
08 · Honest framing

What it will do. What it won't.

It will

Walk you through the full pipeline for a virtual brand: identity, look-dev, motion and voice, distribution, subscription-platform monetization and disclosure, within current platform rules.

It won't
Won't guarantee a brand or audience.
The pipeline organizes production work, but taste, niche selection, consistency, and distribution remain yours. Outcomes vary.
Won't teach you how to game disclosure rules.
We teach you how to disclose properly. AI content + paid promotion needs to be labeled. Doing it right is part of the moat.
Won't pick a niche for you.
Niche selection is in module 1, but the decision is yours. We give the framework; the call is yours.
Won't work for traditional personal brand growth.
If your goal is to be the face, this is the wrong course. Take Instagram Ignited instead.

AI Hook, Bio & Offer Rewriter, included with All-Access.

Paste a hook, bio, caption or offer and get sharper versions written with the frameworks from the courses, in your own voice. Built for creators, not generic marketing copy.

  • Hooks, Instagram bios, captions and offers
  • Five rewrites per run, each with the reason it works
  • Matches the tone of a sample you paste
  • Flags income claims and missing AI disclosure
  • 30 rewrites a day
Open the tool
Sample output from example inputs. Members run it on their own numbers.
Anyro, founder of iimagined.ai
Anyro · founderMelbourne

Meet Anyro, the founder.

Anyro is the founder of IImagined.ai and teaches Instagram growth, AI influencers, digital products and building AI SaaS products. IImagined.ai was founded in 2024 to teach an interconnected system for creator growth, digital products and AI-built software. Anyro also runs a creator management agency, which is where the subscription-platform side of this program comes from.

Read the full story →

Pick how you pay. Get everything.

Both plans open the same thing: all 4 premium courses, the members-only tools, weekly coaching and the community. The courses are not sold one by one.

Monthly

$99/ month

Billed monthly. Cancel anytime.

What you get
  • Instagram Ignited
  • AI Influencers
  • Digital Products
  • AI SaaS Builder
  • Futures Trading course (bonus)
  • Members-only tools
  • Weekly live coaching calls
  • Members-only Discord and the Whop community
  • Every course's resource vault
All-Access, monthly$99/month
Start monthly

Yearly

Best value
$890/ year

Works out at $74.17 a month. You save $298 (25%) against 12 months at $99.

Everything in Monthly: all 4 premium courses, the Futures course, members-only tools, weekly coaching, the community and every vault.

All-Access, yearly$890/year
Start yearly · $890

30-day money-back guarantee on every plan. Cancel anytime; access runs to the end of the period you paid for.

09 · Common questions

What you'd ask before wiring funds.

If your question is not listed, email anyro@iimagined.ai.

All-Access · cancel anytime

Start the work.

AI Influencers and the other 3 courses, weekly coaching and the community for $99/month. 30-day money-back guarantee.

Start All-Access · $99/moOr start with the free Starter Kit →
10 · Next move

What pairs with AI Influencers.

See all 4 programs →
Continue the stack

AI SaaS Builder

Once you can ship a brand, ship the product behind it.

10 listed modules · in All-Access→
Free library

AI Influencers library

Every essay, playbook, and breakdown in the ai influencers category, free to access.

/topics/ai-influencers→
Bundle

All-Access: every program

All 4 premium courses · weekly coaching · private community · resource vault.

$99/mo→
Also exploreInstagram IgnitedDigital Products