Simple-animation accounts: research and approach

Summary

What the research found, in short. Every number is from a secondary source unless it says otherwise.

• Marble races, bouncing-ball loops and treadmill races are the best-known simple formats. Several have large view counts, mostly on YouTube and TikTok.
• One public account races stick figures named by birth month or zodiac sign: @laststickstanding (Last Stick Standing). It is the closest match to Sean's example. Our search found no second account, so the format has only one example and is untested.
• Hooks that appear most in the sources: a payoff visible in the first seconds, an open question, a seamless loop, and a clear rule. Instagram's own ranking rules are not published.
• Brand deals are the main income in the sources, but only after a steady audience. A digital race template or kit is the most likely first product. Platform payouts are small or unpredictable.
• Recommended lead: named-contestant races, 15 seconds or under, looping, with a season table. Figures on this page are planning assumptions to test with our own analytics.

Labels Sourced figures carry the source's own label (Official, Research, Secondary, Unverified, Inference). "Hypothesis" marks our own ideas, which no test has checked.

1. Accounts found

19 accounts across marble races, bouncing-ball loops, treadmill races and object shows. Instagram figures are scarce. Most numbers are YouTube, TikTok or aggregator figures with the date given.

Instagram follower counts were mostly not public in search results. The only Instagram account with a figure is Andreas Wannerstedt, and that figure is from 2019 to 2020. Do not use any number here as a current count.

Gaps and cautions

• Only one account so far races stick figures named by birth month or zodiac sign: @laststickstanding. Our search found no second one, so treat the format as lightly tested. Its views are not public, so its watch-through is unknown.
• No standalone short-form ball-versus-ball account with unique powers was found. Ball Battle (#15) is the closest, and it is long-form.
• A widely shared "automatic ball sorting" video was a Blender simulation, not real footage (Know Your Meme). Label any simulation as one.

2. Hook and retention techniques

Sixteen techniques, with the strength of evidence for each. Instagram does not publish its ranking weights, so most of the platform-level points below come from secondary coverage. Only YouTube's own pages are official.

The "70% three-second retention" and "80% completion" benchmarks conflict across blogs and are not platform-published. Do not use them as targets. No study tested simple animation specifically; the closest work is on ASMR and curiosity.

Sources for this section

• YouTube blog, Sydney Morgan interview on hooks and loops: blog.youtube (fetched 13 Oct 2026)
• YouTube Help, Shorts length and view counting: support.google.com
• Loewenstein (1994), Psychological Bulletin 116(1): 75-98. Barratt and Davis (2015), PeerJ 3:e851: doi. Poerio et al. (2018), PLoS ONE: eprints
• Secondary: Fanpage Karma, Lindsey Gamble, Go-Viral, Social Media Today

3. Monetisation: how these accounts make money

Thirteen methods. Instagram does not publish payout rates, and most figures come from marketing blogs that sell their own tools. Treat every number as a planning assumption.

Planning view: platform payouts (rows 1, 3, 4) are small or unpredictable, so plan on near zero. Brand deals are the main income in the sources. Digital kits can earn more per fan, but sales volume is unknown. No verified income figure was found for this exact format.

How the sell-at-the-end call works

1. Final-frame product reveal. The last second shows the tool or template the animation was made with.
2. One link in bio, to one page. Reels cannot hold clickable links in captions.
3. "Full version or part 2". The clip stops at a cliffhanger, and the full run is on the link.
4. Comment keyword. "Comment RACE for the file." Check Meta's automation and spam rules first.
5. Series. Each post asks viewers to follow for the next round.

Which of these converts best for this niche is unknown. Test one per week and track link clicks.

4. Our approach

A plan for a phase-2 test, built from the research above. Nothing here has been tested, so every choice is a hypothesis until our own test batch shows a result.

4.1 Recommended niche and format

• Lead: named-contestant races. Stick figures or balls, each named for a birthday month or star sign. One winner per run, 15 seconds or under, looping. Why: one account already runs this (@laststickstanding, read 14 Oct 2026), so it is proven only as a format, not as a niche. The closest races hold attention (accounts #2, #3, #12, #14, #18).
• Secondary: ball versus ball, one power each. Ball Battle (#15) is long-form. A short-loop series is a gap in the sources.
• Filler: bouncing and merging balls. The most crowded genre (accounts #8 to #11). Use it to fill the feed, not as the main bet.

4.2 Content plan

1. Week-one test batch. Three formats, a few clips each, in one shared frame layout, so the format is the only thing that changes. Keep it small and fixed in advance.
2. Recurring series with fixed names and numbered episodes. For example a 12-month league, a 12-sign league, and a running season table.
3. Length and loops. 15 seconds or under, ending where it starts. A longer season final can come later.
4. Cadence. A steady weekly rhythm. The research does not give a posting frequency, so set it when we start.
5. Measure per clip. Retention in the first seconds, completion, replays, shares and comments by sign. Our own analytics replace any outside benchmark.
6. Keep or drop rule. Keep a format that beats our own median on those measures. Drop one that does not, after a number of clips set in advance.

4.3 Free tooling

Hypothesis Label every clip as an animation or simulation. We do not claim real footage (difference 10).

4.4 Hook formula

4.5 Monetisation path

1. Build the free audience first. Platform payouts are small or unpredictable.
2. Digital kit at the end of a series. A race template or simulation script sold as a file. Test a small price first.
3. Personalised races on demand. A custom race with the buyer's names. Our idea, not in the research; price not researched (difference 12).
4. Brand deals once there is a steady audience. The main income in the sources. Ranges differ by source.
5. Subscriptions and Gifts only if the account is eligible. Check the Professional Dashboard. Do not plan income on them.
6. Affiliate links and merch as small extras.

4.6 Risks

4.7 Unknowns and follow-ups (not researched)

• Instagram music licensing: which library sounds and trending tracks are licensed for our use.
• Instagram's originality and reuse policy: whether repetitive or re-used content is limited.
• Commissioned-animation rates for brands (monetisation row 13).
• Live follower and view counts for the accounts above. Only Wannerstedt has an Instagram figure, from 2019 to 2020.
• Retention benchmarks: the sources conflict. Use our own numbers.
• Which sell-at-the-end call works best for this niche. Test one per week.
• Eligibility for Gifts and Subscriptions, checked in the Professional Dashboard.

5. Unique differences to stand out

Fifteen ideas for setting the account apart. Hypothesis applies to every item: "none found in public search" means our search did not find one, not that none exists.

1. A fixed roster of named racers. Hypothesis Last Stick Standing (@laststickstanding) already races by birth month and zodiac sign, so the category pick itself is not new. Ours would use the same named racers in every race, so viewers learn each one. Test: replays and comments by racer name, against Last Stick Standing's reels.
2. "Your sign" comment tallies. Hypothesis A pinned tally of comments per sign, updated each episode. Test: sign-prompt clips against clips without one.
3. Season and league tables across episodes. Hypothesis Leagues exist on YouTube, but a season table on a short-form feed was not seen in the sources. Test: do viewers who see the table come back?
4. Recurring characters with one trait per sign. Hypothesis Each sign has a fixed trait that shows in races. Test: do viewers name a favourite in comments?
5. Ball-versus-ball powers as a short loop series with a visible rulebook. Hypothesis Ball Battle is long-form; no short-form version found. Each episode repeats the rule card and adds one power. Test: completion with and without the card.
6. Rule card on screen in the first second. Hypothesis Short rule text in frame 1, readable with sound off. Test: first-second hold with and without it.
7. Seeded runs viewers can ask to re-run. Hypothesis Every run has a seed; viewers ask for a re-run with a new seed. Test: re-run requests per clip.
8. Viewers pick the winner before the run. Hypothesis Pause, ask "pick a winner", then show the result. Test: completion and comments with and without the prompt.
9. Original sound for each event. Hypothesis Our own generated notes, one per event. Trending-song covers are used by some accounts. Test: replays with and without sound.
10. An honest "simulation" label. Hypothesis Every clip says whether it is an animation or simulation. Test: questions in comments about whether it is real.
11. Instagram-native vertical framing and safe zones. Hypothesis Most sources are YouTube or TikTok. Test: how text sits on a phone screen in the test batch.
12. Personalised races on demand. Hypothesis A custom race with the buyer's names, delivered as a video. Price not researched. Any sale needs Sean's approval.
13. A digital kit at the end of a series. Hypothesis Template packs exist (monetisation #8), but none for the race format was found. Test: one listing price before building more.
14. Tournament brackets across the week, ending in a final. Hypothesis Tournaments exist on YouTube. Test: do the finals get more replays and shares than weekday clips?
15. Accessibility: colour-blind-safe palettes and large labels. Hypothesis Rarely seen in the sources. Test: check the palette with a colour-blindness simulator before the batch.

Last Stick Standing (@laststickstanding)

The closest public match to the named-racer idea so far: stick-figure races where the viewer picks their own category (birth month, zodiac sign, name initial, country). Researched 14 October 2026 from public pages only. Its 13 Oct reel and five YouTube Shorts were watched on 14 Oct 2026 (T-29), so the race structure is now Observed; the other Instagram reels were read as captions and counts only. Instagram views are not visible.

Correction Earlier statements on this page that no stick-figure race named by birth month or star sign exists (Summary, Gaps and cautions, and difference 1) were written before this account was found. This account runs birth-month and zodiac races, so treat those statements as superseded.

Account facts

Recent reels and reach

Reach: the 10 Oct reel has 116.9K likes on 2,121 followers, about 55 times the follower count (our calculation). Most reach therefore comes from the recommendation feed, not from followers. Comments are about 3% of likes on that reel and about 6% on the 11 Oct reel (our calculation). Views are unknown, so watch-through is unknown.

Format, hook, pacing and staging

What we copy

1. A question hook that names the viewer's own category ("Which sign wins?"). Official pattern on the top reels. Whether it drives comments is Inference from the comment counts.
2. One repeatable template, with a theme swap per post (castle edition). Unverified from one caption.
3. A numbered series ("Part 1"), so there is a reason to watch the next one. Official caption.
4. A clear finish order at the end, so the viewer can find their pick. Unverified.
5. Original audio, not licensed music. Official label on one reel.
6. One hashtag set plus one account tag. Official hashtags; effect Unverified.

What we differ on

1. Show the winner's name in the first second, not only in the caption. Hypothesis Untested; the account's frame 1 is not known.
2. A one-line reason for each entrant, for example a birthday-month trait, so viewers care about a racer and not only their own pick. Hypothesis
3. End on a choice for the next round ("rematch with your sign"), not only a comment. Hypothesis It reuses the loop.
4. A fixed length and a fixed racer count (for example 15 s, with the same number of racers), so the format is easy to repeat and test. Hypothesis
5. Sell one thing at the end. No product is visible on this account. Hypothesis A test in phase 2 only; any sale needs Sean's approval.
6. Separate themes per category family (months, signs, countries), so one audience is not split across them. Hypothesis

Open items

• Watch the 10, 11 and 12 Oct reels with sound on, and note the loop and the first 3 s on mute. The 13 Oct reel and the Shorts are done (T-29).
• Read the YouTube channel's subscriber count in a browser. The fetch returned none.
• View counts are not public. They need the account owner's stats or a public stats site.

Sources

• Instagram profile (read 14 Oct 2026, Official). 13 Oct reel, 12 Oct reel, 11 Oct reel, 10 Oct reel (captions and counts, Official). The 9 Oct reel's fields were not readable.
• YouTube channel (linked from the bio; no figures returned). No Secondary sources were found for this account.
• Full notes: pm/research/laststickstanding.md.

Sample video

Two test clips, both rendered by our own code (render_race.py). Nothing is posted or for sale.

Hazard race sample: foods

Six everyday foods race side on while the camera scrolls with the pack. Every few seconds a wave of hazards flies in: the other racers jump or duck, one racer is hit and is out, and the last food left wins. Mode elimination_v2 (version 2 of the hazard race), 26 s (24 s of race, then the winner card), 1080x1920, with sound. Each racer has its own look (dog, fish, robot, round, square, stick) and each wave is named in plain words at the bottom, for example "ARROW BLADE". The hazard idea follows the competitor races in the T-29 tables below; the foods, looks, lanes and hazards are drawn by our own code, with no template and no watermark. Direct link: sample-race-hazards.mp4; config: sample-race-hazards.json.

Who is in the race and what knocks them out

Observed Frames at 7.5, 12, 15, 18 and 22.5 s show "SUSHI OUT" (5 LEFT), "FRIES OUT" (4 LEFT), "BURGER OUT" (3 LEFT), "DONUT OUT" (2 LEFT) and "PIZZA OUT" (1 LEFT); the finish line is in frame at 22.5 s. Frames at 24.2 and 25.5 s show "TACO WINS!" and the call to action. Unverified The exact knock-out times and which hazards each racer dodged come from the render plan, not from a frame check. A person still needs to watch the clip.

Automatic check (verify_video.py): the checker does not know version 2 yet, so it treats this clip as a 15 s looping race. It reports 6 of 8 hard checks passing; the 2 fails are "15 s or under" and "last frame matches first", which do not apply to a race that ends on its winner. Teaching the checker version 2 is a fix still to make in the skill.

What the competitor races do (T-29, read from video)

Observed seen in full frames or clearly readable frame tiles. Unverified seen once on a contact sheet, or inferred. Times are about ±2 s and ±1 racer. Hazard names are our labels. Source: pm/research/T-29.md, read 14 Oct 2026.

• Observed 9:16 vertical, about 54 s, 12 racers per race. The side-on track runs diagonally and the camera scrolls down with the pack; it does not zoom out.
• Observed Beats: hook card "PICK YOUR ..." with "ONLY ONE SURVIVES" (0 to 2 s), countdown 3, 2, 1 (2 to 4 s), start (4 to 5 s), eliminations (about 8 to 47 s), the last three reach a FINISH gate (47 to 50 s), then the winner's name and confetti to the end (50 to 54 s).
• Observed On screen: "N LEFT" counter top-left; "[NAME] OUT" banner with a skull icon top-right; "[NAME] LEADER DOWN" when the leader falls.

Racers left by time

Observed counter reads, about ±1 racer. The month Short and the 13 Oct reel share the same sequence and winner; they are not pixel-identical (Unverified).

Hazards and how they knock racers out

Observed knock-out mechanics: H1 to H4 only. Every other hazard is a shape seen on one contact sheet and needs a frame check before we copy it.

• What we take: the same beat sheet (hook card, countdown, one knock-out every 3 to 5 s, FINISH gate, winner to the end). Hypothesis Our clip keeps it in 26 s: 24 s of race and 2 s of winner card.

Stick-figure race sample

Second clip, same skill, style: "stick": 12 star signs run as stick figures (Castle Edition theme), 15 s, 1080x1920, one winner, one CTA. Staging follows the Last Stick Standing race structure read from video in T-29 (hook card, countdown, one knock-out every few seconds, finish order at the end). Our frame 1 and pacing have not been checked by a person yet. Direct link: sample-stick-race.mp4; config: sample-stick-race.json.

6. Next steps: a 30-day test plan

A plan for phase 2, if Sean approves it. Nothing here is built, posted or bought yet. Every number is a planning assumption until our own analytics show a result.

6.1 Chosen niche

• Lead: named-contestant races. Stick figures named by birthday month or star sign, one winner per run, 15 seconds or under, looping. Basis: section 4.1.
• The only public example we found is @laststickstanding, 2,121 followers Official, read 14 Oct 2026. Its race structure was read from video (T-29); its Instagram views are not public.
• Secondary: ball versus ball, one power each. Filler: bouncing and merging balls.

6.2 First 10 video ideas

Hypothesis Titles, hooks and twists are our own ideas, not tested. Hook technique numbers refer to research/hooks.md.

6.3 Posting cadence

• One clip a day for 30 days, at one fixed time (18:00 local, Hypothesis). A fixed slot keeps timing out of the test.
• Days 1 to 10: ideas 1 to 10, in order, one each. Days 11 to 30: the name set and format that did best, with a new twist each day.
• Keep one frame layout for the whole test (section 4.2), so the format is the only thing that changes.
• Label every clip as an animation or simulation (difference 10).

6.4 Free tools only

• Render engine: the root-for-yours skill in skill-src/root-for-yours. Start from examples/config-stick-star-signs.json and render with python3 scripts/render_race.py <config> out.mp4.
• Checks: scripts/verify_video.py runs the automatic checks. A person then watches each clip against the techniques checklist.
• Other: Python with Pillow and numpy, ffmpeg, and original or CC0 sound only.

6.5 End card and monetisation order

Each end card has one ask, not three. The order follows section 4.5.

6.6 Success measures after 30 days

Keep or drop Keep a format that beats our median on at least two of the first four measures. Drop a format that does not, after 10 clips.

6.7 Open items

• Sample audio needs a human listen. The samples use synthesized tones. Items A1 and A2 in pm/research/T-20.md stay open until someone listens.
• 11 external links block bots. A browser check is needed before any link in this report is relied on.
• Phase 2 needs Sean's approval: creating an account, posting, and any spend.
• Not researched: payment method and fees for a kit.

Sources and notes

• Research files in the project folder: research/accounts.md (19 accounts), research/hooks.md (16 techniques), research/monetisation.md (13 methods), research/approach.md (plan, differences and follow-ups).
• Public web only. No Instagram login, no scraping, nothing posted. Instagram Help Center pages returned only the home page, so Meta claims are marked secondary.
• Nothing here was checked on live platforms. Re-check any figure before a decision.