BusinessRnD · draft v2 · built 2026-10-21 06:15 UTC
I7 · Owned 2D character universe — deep dive
Skip
A lottery ticket: fun and cheap, not a plan.Realistic: ~$900/mo expected at month 24 · Chance of $10k/mo: ~3%
Cards have Go deeper, Maybe and Kill buttons. Press Send feedback to save. Your marks from all tabs collect at the end of each page.
Prepared 2026-10-20. Counts from public YouTube and Graphtreon pages were read the same day. "Estimate" marks my own arithmetic. Full evidence is in the research notes.
Bottom line
Being AI-made isn't what gets channels killed. Looking like slop and repeating one template is. Clearly unrealistic animation needs no AI label. YouTube says a label "alone won't limit" reach or earnings, and its rules allow "a series following a set of characters" (YouTube). In January 2026, YouTube banned or emptied 16 AI channels with about 35M subscribers for templated sameness (OutlierKit).
The honest benchmark is small.
The best-known adult AI character universe, Neural Viz, has 238K subscribers and about 15M lifetime views after 28 months (YouTube).
The AI character channels with billions of views are Shorts-first and appeal to kids, and their income figures are third-party models.
No solo AI character channel has a verified $10k/month.
Your edge is cost and consistency. Characters rigged in code put a 10-minute episode at about $3–6 in generation fees and 6–11 hours of AI agent time (estimate). Generating every shot with a video model costs about $35–60, and the characters drift between shots.
The format decides the money.
Explainers earn about 4× the ad income per 1,000 views of comedy or story ($10.22 vs $2.43 median — AIR via Clipflux).
Serialized stories turn fans into paying members best: 0.9% of subscribers for the MonkeyWrench web series, against 0.02–0.1% for most animation channels (Graphtreon).
Two things break "your recurring work ≈ 0":
Copyright needs a human-written layer; approving AI output isn't enough (USCO).
Budget 2–4 hours once for a canon in your own words, plus about 15 minutes a month if you use Patreon.
Outlook (estimate):
About a 3% chance of $10k/month by month 24, and about a 55% chance the pilots are killed within 90 days.
Expected revenue at month 24: about $900/month.
A 3-month test costs about $150–700 plus 5–8 of your hours. Take it as a cheap option, not as cash.
Is being AI-generated a problem?
Short answer:
Policy: no, if you avoid sameness and kid appeal.
Perception: yes, if it looks like slop.
Copyright: yes, unless you add a human layer.
Risk
Verdict
Evidence
What we do
AI label
Low
Unrealistic animation is exempt, and a label doesn't cut reach or pay (YouTube). Labels are applied automatically to photoreal clips, files with C2PA metadata and clips made with Google's tools; labels on Veo clips can't be removed (Tubefilter)
Animate in code, which leaves no AI metadata. Use generated clips rarely. One plain "made with AI tools" line in the description
AI music as the main focus
Medium
It needs a label in the player (YouTube). Spotify demotes AI personas (Spotify)
Songs only as theme tunes, not a music channel
Templated, mass-produced content
Medium, controllable
Policy bans "generic or unoriginal templates". Channels whose stories "differed only in names and plot fragments" were banned in Jan 2026 (OutlierKit; YouTube)
Weekly, not daily. Varied episode structures, season arcs, unique thumbnails, and an automatic sameness check
Made for kids
Medium–high
Cartoon characters are a listed factor. YouTube can override your setting and lock it. Kids' videos lose comments, memberships, personalized ads and merch (YouTube; YouTube). Kids RPM is $0.33
16+ themes and vocabulary, no toddler colours, no cute baby animals, no nursery songs
Teen viewers
Medium
US users estimated to be under 18 get no personalized ads (YouTube)
Aim at 18–34 and count teen views as lower-paying
Audience mood
Medium–high in Western adult niches
"Slop" was the 2025 Word of the Year (ADS). Aronofsky's AI series was panned as "slop" (THR). 39% of Gen Z feel negative about AI (IAB via Tubefilter). Yet labelled AI characters still draw mass views. No data on how labels affect retention
Hand-made look, real jokes, good voices. Track the share of "AI slop" comments as a kill trigger
Sponsors
Mixed
A labelled AI Short with "Includes paid promotion" got 49.6M views in Sept 2026 (YouTube). McDonald's Netherlands pulled its own AI ad after backlash (Forbes). 73% of consumers say disclosure helps or doesn't hurt buying (IAB)
Expect mass-market and AI/software sponsors. Deals need your signature
Copyright, licensing, merch
High unless fixed
Prompts alone don't create authorship (USCO). The Supreme Court declined the AI-authorship case in March 2026 (docket). The owner of Tung Tung Tung Sahur fell back on trademark (GamesBeat)
You write a one-page canon and each season's story beats. We keep an authorship log and trademark the series name once a pilot passes
EU AI Act
Low
From 2 Aug 2026, fictional works only need disclosure "that does not hamper the display or enjoyment" (Art. 50)
The description line covers it
Pros and cons
Pros
Cons
You own the channel, the audience and the trademarks. AdSense and Patreon both pay a Ukrainian FOP
No verified solo AI character channel at $10k/month. Neural Viz is small
Code-rigged characters: about $3–6 per episode in generation fees, and identical in every frame
Writing is the bottleneck, and AI is weakest at comedy
The AI runs script, render and upload through official APIs
Western adult and teen viewers dislike "slop"
Policy allows character series, and cartoons need no AI label
Copyright is thin without your written layer
The back catalogue keeps earning; YouTube auto-dubs into other languages; licensing upside
Slow: 6–12 months to reach the monetization bar, which doubles on 1 Feb 2027
Several income streams: ads, members, sponsors, merch, a Ukrainian Cultural Fund (UCF) grant
Being reclassified as kids' content, plus teen viewers, cuts ad income. Mass reporting and copycats
Reuses your Spine, story-engine, speech and mix tools
Licence cleanup first: Spine licence, Vortex art, aligner, music terms. Patreon can't be posted to automatically
quadrantChart
title Universe options by safety and upside
x-axis Low policy and audience safety --> High policy and audience safety
y-axis Low revenue upside --> High revenue upside
quadrant-1 Build first
quadrant-2 Upside but risky
quadrant-3 Avoid
quadrant-4 Safe but small
Lab Crew: [0.78, 0.72]
Signal Lost: [0.66, 0.62]
Agent Office: [0.6, 0.47]
Folklore Night: [0.56, 0.3]
Music band: [0.36, 0.28]
Brainrot Shorts: [0.14, 0.42]
Kids cartoon: [0.18, 0.12]
Risk of being treated as reused content. No AI "experts" on health, legal or money
Shorts earn 3–14% of the long-video rate (AIR). Use them to bring viewers to the full episode, not as income.
Cadence: weekly 8–15-minute episodes plus 3–5 Shorts cut from each. From 2027 that easily meets YouTube's rule of 2 long videos plus 5 Shorts every 90 days to stay active.
Facebook pays Ukrainians in an invite-only beta. You need 5–10K followers, and one creator got $97 for 1.3M views (Cityhost). Cross-post for reach only.
Membership math: 500 members at about $6 needs about 55K subscribers at the serialized-story rate, or 0.5–3M at typical rates. Patreon takes 10%.
Production pipeline
flowchart TD
A["Season arc and canon - you approve once per season"] --> B["Episode script from your story engine"]
B --> C["Canon, sameness and kid-appeal check"]
C --> D["Voices via ElevenLabs on Kie - about 1.60 USD"]
C --> E["New backgrounds and props - about 1 USD"]
D --> F["Word timings to mouth shapes - free"]
E --> G["Spine rigs animated in code - free"]
F --> G
G --> H["Optional generated clips - 0 to 0.80 USD"]
H --> I["CPU render, captions, mix"]
I --> J["Quality check and repairs"]
J --> K["Thumbnail and title tests - about 0.25 USD"]
K --> L["Upload via YouTube API, set not made for kids"]
L --> M["Cut 3 to 5 Shorts, copy to Facebook"]
M --> N["Metrics review feeds the next script"]
Per 10-minute episode (estimate, from unit prices measured in your projects):
Cost:
About $3–6 in generation fees: Kie is $5 per 1,000 credits, an image is 8 credits, and video is 19 credits per 5 seconds.
ElevenLabs costs about $0.06 per 1,000 characters (Kie).
The same length made with a video model: about $11 per take on Seedance (about $35 with typical retakes) or about $55 per take on Higgsfield 480p, with characters drifting between shots.
Agent time: about 6–11 hours. Script 1–2 h, animation 2–4 h, quality check 1–2 h; rendering runs at about real time on CPU.
One-time costs:
Spine licence in your name: $69 or $379 (Esoteric). Your tools use Spine's own software components, which require an editor licence.
Character art: about $5–20.
An optional ElevenLabs plan at about $22/month.
Where your taste is needed:
Once: cast, voices, style, and the canon in your own words.
Each season (8–10 episodes): approve the arc, 30–60 minutes.
Escalations: a kids'-content override, policy flags, sponsors.
Licence flags:
Spine2D-Rnd was built for Vortex with Vortex art (Question 3).
Swap out the non-commercial word aligner your tools currently use (MMS_FA) before launch.
Check Suno's terms, or use YouTube's royalty-free music tool, which is in testing (Tubefilter).
Characters and voices must be fully synthetic: no visual-passport likeness, no voice clones.
I7.1
Lab Crew
What it is: Character-hosted science, tech and history explainers. A cynical robot and an overeager lab rat take apart one real question per episode, with 16+ humour.
Audience: Curious adults aged 18–34 in the US, UK, Canada and Australia.
Format and cadence: 8–14-minute episodes weekly, plus 4 Shorts each.
SolarBalls (character-led space education) has 2.27M subscribers and 2,095 patrons (Graphtreon).
Kurzgesagt takes 65% of its income from viewers (Wikipedia).
Why we can win: Your research scrapers plus Claude's strength at explaining. Consistent characters cost almost nothing per episode.
Monetization stack: Ads, then sponsors, Patreon topic votes, and later merch.
Main risk: Cute characters get reclassified as kids' content. Accuracy errors. Overlap with I9, so pick one channel.
First step: Pilot 1 episode and 10 Shorts on one sub-topic.
I7.2
Signal Lost
What it is: A serialized sci-fi mystery about a night-shift crew on a relay station. Every episode ends on a cliffhanger.
Audience: Fans of animated drama and lore, aged 16–34.
Format and cadence: Seasons of 8 episodes, 10–12 minutes each. Vertical recaps go into YouTube's "Shorts Series" playlists.
Evidence it can work:
MonkeyWrench converts 0.9% of subscribers into paying members at about $6 each.
The Amazing Digital Circus pilot drew about 441M views, then a Netflix licence and a cinema release (Wikipedia).
Why we can win: Your micro-drama story engine writes hooks, and your continuity tools plus Spine keep 50+ episodes consistent.
Monetization stack: Patreon early access and lore, ads, later licensing.
Main risk: Viewers drop between episodes, new viewers must start at episode 1, and the English dialogue may not be good enough.
First step: Pilot episode 1 and 10 Shorts, then measure how many return for episode 2.
I7.3
Agent Office
What it is: A mockumentary about a startup run by AI agents and one exhausted human founder.
Audience: Tech workers and AI-curious adults.
Format and cadence: 6–10-minute episodes weekly. Cold opens become Shorts.
Evidence it can work: Neural Viz's mockumentary formats reached 238K subscribers (YouTube). AI and software tools are natural sponsors.
Why we can win: You live this material (agents, worktrees, quality gates). Saying openly that it's made by agents turns the AI liability into the premise.
Main risk: Comedy is the hardest thing to write. It could come across as self-promotion.
First step: Keep it as the backup pilot if you prefer comedy.
I7.4
Folklore Night
What it is: A Slavic folklore horror-comedy anthology, hosted by a sarcastic house spirit (domovyk).
Audience: Global horror and folklore fans, plus a Ukrainian audience through a dub.
Format and cadence: A 10-minute standalone tale every two weeks.
Evidence it can work: The Ukrainian Cultural Fund (UCF) funds "animated series for social networks" with 500K–3M UAH grants, and FOPs are eligible (UCF). No demand data was found for this niche.
Why we can win: Ukrainian authenticity and grant eligibility.
Monetization stack: Grant, ads, later book or game licensing.
Main risk: Horror limits which ads run. The niche is small. UCF's view on AI is unknown.
First step: Check the dates of UCF's next call. Pilot only if a grant looks reachable.
Producing your own content is allowed operator work. No API credits
Character sheets and layered art
Existing non-Anthropic gen tools
You (Kie, or image generation in Codex on your ChatGPT plan)
One-time, $5–20. Fully synthetic characters
Backgrounds, props, thumbnails
Existing non-Anthropic gen tools
You (Kie, about $0.04 per image)
About $1–1.50 per episode
Character voices
Existing non-Anthropic gen tools
You (ElevenLabs via Kie)
Designed voices, not clones
Lip-sync timings, animation, render, mix
No model
You (server already paid)
Runs on your server's CPU
Optional video clips, theme song
Existing non-Anthropic gen tools
You (Kie Seedance or Suno, Higgsfield)
Clips from Google models carry labels you can't remove. Check Suno's terms
Uploads, Shorts, Facebook copies
No model
You
YouTube Data API, Meta Graph API
Extra-language dubs
No model
YouTube
YouTube's auto-dubbing runs on its side
Anything viewers trigger
No model
Nobody
Nothing runs when people watch
Next steps
gantt
title I7 test plan from 21 Oct 2026
dateFormat YYYY-MM-DD
axisFormat %d %b
section Setup
Licences rigs and publisher :s1, 2026-10-21, 10d
section Pilots
Produce two pilots :p1, 2026-10-28, 12d
Publish pilots and Shorts :p2, 2026-11-09, 21d
Day 21 go or kill :milestone, m1, 2026-11-30, 0d
section Series
Weekly episodes if go :e1, 2026-12-01, 61d
Old YPP bar ends :crit, milestone, m2, 2027-01-31, 0d
Day 90 review :milestone, m3, 2027-02-07, 0d
section Scale
Members and dubs if monetized :x1, 2027-02-08, 56d
I do:
Build your character pipeline: synthetic layered art → Spine rigs → mouth shapes timed to the voices → render. It reuses your Spine2D, story-engine, continuity and mix tools, plus the missing YouTube publisher.
Write the bibles and scripts for two pilots, I7.1 and I7.2. Make 1 episode and 10 Shorts for each, publish them over 3 weeks and reply to comments.
Check every upload for sameness, kid appeal and policy. Measure the real cost per episode.
Report the day-21 numbers with a go/kill call.
You do once:
Pick the pilots, or give me your own concept. Approve the cast, voices and style (1–2 h).
Write or dictate a one-page canon in your own words (1–2 h). This is your copyright layer.
Buy a Spine licence in your name, and confirm whether the Spine2D-Rnd code is yours (Question 3).
Set up YouTube (shared with I9): phone verification, 2-Step Verification, and the API connection set to stay logged in.
Later, at the monetization bar: AdSense and the W-8BEN tax form, Patreon via Payoneer, a trademark on the winning series.
You do recurring:
Optional: approve each season arc (30–60 min per 8–10 episodes).
If we use Patreon: paste the monthly batch of posts I prepare (about 15 min).
Sign any sponsor deals.
Go/kill thresholds:
Day 21, per pilot. Go if all four hold:
at least 1 subscriber per 1,000 Shorts views;
long episodes watched for 40% or more of their length on average;
at least 5% of comments name a character;
viewers come back for episode 2.
Kill if, after about 500K Shorts views, there are fewer than 0.3 subscribers per 1,000 and nobody names a character.
Any time:
A limited-ads flag, a templated-content flag or a kids'-content override: pause and fix before the next upload.
More than 10% of top comments call it "AI slop": change the look or the voices.
Day 90: at least 1,000 subscribers and 1,500 watch hours, or kill.
Month 6: eligible for monetization, or kill.
Month 12: at least $1,000/month from ads, members and sponsors combined, or keep it only as a hobby.
Risks and open questions
Writing quality in English comedy and drama is the biggest unknown. Only the pilots can answer it.
Authorship: how much will you write yourself? Without it, licensing rests on trademarks alone.
Ownership and licences:
Spine2D-Rnd is Vortex code with Vortex art.
The MMS_FA word aligner is non-commercial.
Suno and Eleven Music limit commercial use of their output.
The sound-effects library needs Freesound credits.
API gaps:
Patreon posts and YouTube community posts can't be published through the APIs.
Members-only uploads and extra audio tracks through the API are unverified.
Uploads from an unverified API project may be locked private; we test once.
Unknowns in the data:
How much teen viewers cut ad income.
Whether AI labels affect retention; no data found.
No verified revenue for any AI character channel.
Mass reporting and copycats: Fruit Love Island died from mass reports. Publish on YouTube under trademarked names.
Overlap with I9: Lab Crew could become I9's look. Choose one channel, not two.