STUDIO / PROJECT 001CREATIVE DEVELOPMENT ●

THE CREATIVE NORTH STAR

Make something
worth watching.

Stories, characters and moments you want more of.
Built with AI. Directed by taste.

Explore the creative slate ↗Download the brief ↓
01 / COMEDYEveryday life.
Extraordinary cats.
02 / CINEMAA triumphant
return to Mars.
03 / REMIXGood ideas.
New audiences.
04 / CREATORA personality
worth following.
100Kviews on one video
1Kfollowers · fresh account
90 dayslaunch date to be set
Enjoy it.the reason to keep going
01 / INTENTION

A creative practice that can become a media business.

Create short-form entertainment I would enjoy watching, scratch my creative itch, and learn to produce it programmatically with AI. Build recurring formats with the potential for serialization, broad reach, and eventual monetization. The audience should care about the story, humor, characters, or usefulness—not the fact that AI helped make it.

Enjoyment and sustained creative practice matter alongside audience growth. Avoid generic, visibly low-effort AI output.

90-day target: one video reaches 100,000 views and a fresh account reaches 1,000 followers. The launch date is not yet set; these are targets, not forecasts.

Creator is based in California, willing to use his own face and voice, and can commit 30–40 hours a week if the work stays interesting.

Production budget is undecided. Ramp up with small experiments before committing; an unset budget is not unlimited spending authority.

Primary distribution: TikTok, YouTube Shorts and X. Explore language localization after a format works. Existing audiences do not have to be used.

Use interests revealed by selected X bookmarks and TikTok favorites as research clues. Shared YouTube history includes children’s viewing and must be filtered with Eng.

Manage work through AI chats; keep a shared command page as the persistent reference. Preferred home is engtaing.com on Cloudflare, with a shareable brief and gated working area/assets.

Current stage: discovery. No pilot has been produced, no fresh-account results are recorded, and no production budget has been selected.

02 / THE SLATE

Four directions. One first experiment.

The cat sitcom is the proposed starting point. The other tracks remain available; a recommendation is not a committed production order.

01Recommended first experiment

The cat-couple sitcom

20–45 sec · recurring comedy

Inspired by the Chinese cat-couple videos Eng’s girlfriend enjoys on Douyin, WeChat and Rednote. Explore an English/American version with everyday relationship humor.

First proof

Write three original domestic-comedy premises. Choose one, make a six-shot storyboard and a short performance test before a complete episode.

Approach, reusable pieces & open decisions

Approach

Create original recurring cats, voices and domestic settings; or seek permission to localize the source creator’s footage. Adapt comic situations and timing for the audience. Exact footage, scripts and audio are not assumed cleared for reuse.

Reusable production pieces

Character sheets, sofa/kitchen sets, voice identities, reaction shots, subtitle styles and shot/edit templates.

Still to decide

Original series or licensed localization? Which character dynamic and English voice pairing feels funniest to Eng and his girlfriend?

What success looks like

One coherent episode with stable characters, readable action, natural dialogue, a clear punchline and a measured cost/time record; Eng wants to see episode two.

02Scene clarification needed

Red Rising, brought to life

1–2 min · cinematic / animated

Visualize the moment the main character descends on Mars triumphant. Test photoreal or animated cinematic production around a book Eng loves.

First proof

Confirm the scene, then propose two visual treatments and a 10–15 second proof of atmosphere, movement and scale.

Approach, reusable pieces & open decisions

Approach

First identify the exact book and moment. The Iron Rain on Mars and a later victory procession are possible interpretations, not confirmed selections. Plan a brief look-and-motion test before a full scene.

Reusable production pieces

Art direction, environment references, camera language, sound palette and a consistent character-reference workflow.

Still to decide

Exact scene/chapter, desired fidelity, style and rights route for publication or monetization.

What success looks like

The chosen scene is documented accurately; a short test conveys the intended emotion without treating unverified canon or publication rights as settled.

03Research track

Ideas worth translating

Short-form · multilingual

Find English content Eng genuinely enjoys and develop concise versions for English and other languages.

First proof

Select five examples, explain their appeal, confirm which reflect Eng’s viewing, and pitch three original adaptations.

Approach, reusable pieces & open decisions

Approach

Separate a reusable format or broad idea from protected footage and expression. Produce original scripts and visuals, or use authorized source material. Localize jokes and context, not just words.

Reusable production pieces

A format library, hook patterns, language-separated dialogue tracks, subtitles, edit timing and localization review.

Still to decide

Which niches and languages? Which YouTube items are Eng’s rather than his children’s? No YouTube-history review is complete.

What success looks like

One original concept works in English; a second language preserves the payoff and is reviewed by a competent speaker.

04Exploratory

A virtual creator with a point of view

15–45 sec · fictional adult host

Create an attractive fictional adult female personality around interests Eng enjoys: potentially lifestyle, products, gadgets, food, travel or humor.

First proof

Develop a character brief, three content pillars and three episode ideas. Test identity consistency only after Eng chooses the direction.

Approach, reusable pieces & open decisions

Approach

Define personality, taste and recurring situations before optimizing appearance. Use a consistent original identity, avoid false real-person testimonials, and use appropriate synthetic-media and sponsorship disclosure.

Reusable production pieces

Identity references, wardrobe, voice, locations, product-shot templates and a repeatable editing style.

Still to decide

Which audience and point of view? Entertainment, product demonstrations or eventual advertising? No identity or voice has been created.

What success looks like

Viewers have a reason to return beyond appearance; identity holds across shots and any factual/product claim is supportable.

03 / PROJECT MANAGEMENT

The next useful move.

12 tasks
MC-01 / StudioReady for review

Capture the creative brief and shared project state

Next: Eng reviews this first revision; corrections become a new version.

Owner & acceptance

Owner: Codex

All four tracks, constraints, references, open decisions and handoff are captured.

MC-02 / StudioDone

Publish on the owned domain and protect the working area

Next: Public brief, protected workspace and private source are verified. Future updates should keep these three surfaces on a recorded revision.

Owner & acceptance

Owner: Codex

Public brief opens without login; private working content is denied to an anonymous visitor.

MC-03 / CATDecision needed

Select the first production experiment

Next: Review the cat sitcom recommendation against the other three tracks.

Owner & acceptance

Owner: Eng

One pilot, audience, approximate length and spending limit are chosen.

MC-04 / CATIn progress

Break down the source comedy and localization opportunities

Next: Complete dialogue/voice analysis of the selected examples; separate observed beats from inference.

Owner & acceptance

Owner: Codex

A beat sheet records setup, escalation, punchline, timing and what depends on Chinese language/culture.

MC-05 / CATQueued

Create three original pilot premises

Next: After MC-03, write hooks, beats and endings; Eng selects one.

Owner & acceptance

Owner: Unassigned

Each premise stands alone, suits recurring characters and has a distinct payoff.

MC-06 / CATQueued

Build the character and performance proof

Next: After MC-05 and a generation budget, test consistent references, voice and a short animated exchange.

Owner & acceptance

Owner: Unassigned

Identity, expression and dialogue hold across shots; rejected outputs and costs are recorded.

MC-07 / REDDecision needed

Confirm the Mars scene

Next: Identify book/moment and cinematic versus animated preference.

Owner & acceptance

Owner: Eng

Scene selection is explicit; any adaptation-rights question remains recorded.

MC-08 / REMIXQueued

Build a taste-led reference shortlist

Next: Use selected bookmarks/favorites; ask Eng to resolve ambiguous shared viewing.

Owner & acceptance

Owner: Unassigned

Five entertainment references with reasons to watch and three original concepts.

MC-09 / UGCQueued

Define the fictional host and recurring premise

Next: Offer three personality/content combinations before generating identity assets.

Owner & acceptance

Owner: Unassigned

Adult identity, audience, voice, recurring content and disclosure approach are clear.

MC-10 / StudioQueued

Compare director models on the same brief

Next: Provide this revision to Codex, DeepSeek harness and the selected Claude setup.

Owner & acceptance

Owner: Eng + chosen models

Same inputs, task, budget and rubric; separate model performance from tool/harness access.

MC-11 / StudioQueued

Finish, review and publish the first episode

Next: After a pilot passes creative review, prepare platform exports and obtain publication instruction.

Owner & acceptance

Owner: Unassigned

Finished vertical video, captions, cover, rights notes, disclosure decision and cost/time record.

MC-12 / StudioQueued

Measure and decide what deserves a second episode

Next: Review performance at consistent intervals after publication.

Owner & acceptance

Owner: Unassigned

Record reach, retention where available, shares, follows, production effort and Eng’s enjoyment; choose iterate/continue/stop.

Dated planning snapshot. Task status changes are made through an authorized AI update and a new published revision; this page does not dispatch agents or synchronize chat activity automatically.

04 / THE PRODUCTION SYSTEM

Repeat the craft. Keep the judgment.

  1. 01

    Discover

    Collect a small reference set and articulate why Eng would watch.

  2. 02

    Write

    Choose an audience promise; write hook, escalation, payoff and next-episode potential.

  3. 03

    Design

    Lock references, characters, sets, shot list and a versioned episode plan.

  4. 04

    Produce

    Generate or film short shots; keep accepted and rejected takes with settings and costs.

  5. 05

    Assemble

    Edit programmatically where useful; keep dialogue, music, effects and captions as separate tracks.

  6. 06

    Review

    Check story, pacing, consistency, rights, audio and any disclosure; obtain Eng’s creative review.

  7. 07

    Localize & publish

    Adapt language/culture from the approved master; export for each platform and publish only when instructed.

  8. 08

    Learn

    Log available retention, shares, follow conversion, cost/time and enjoyment. Change one creative variable at a time.

Proposed first sprint · start date unset

Week 1: select one pilot, develop three premises, choose a look and test a performance. Week 2: assemble one complete episode, review it, and prepare distribution. Adjust the schedule to the evidence; do not batch weak episodes just to fill a calendar.

05 / DECISION LOG

Know what is settled. Know what is not.

Confirmed

The audience experience comes first

AI is the production method. The chosen content should be enjoyable in its own right.

Confirmed

Start with a fresh growth target

One 100K-view video and 1K followers in 90 days; launch date remains unset.

Confirmed

Owned-domain command center

Eng prefers engtaing.com on Cloudflare: shareable brief, gated working area/assets, operated through AI chats.

Proposed

Test the cat sitcom first

Low scene complexity, recurring characters and a short joke make production learning easier. Eng has not yet selected or funded this pilot.

Proposed

Prove a performance before scaling

One strong scene before a batch of episodes; character and timing failures are costly to repeat.

Open

Budget, rights and exact production scope

Choose a per-pilot spending ceiling, source-material route, first language and first account.

Open

Red Rising scene

“Descends on Mars triumphant” is not yet specific enough to lock a canon scene.

06 / REFERENCE ROOM

Study why it works.

REFERENCE ↗

我哩个豆 · Episode 37 · 套路 / The setup

28 seconds. Orange-and-white and calico cats on a sunlit sofa; money/gift setup and a physical reaction payoff. Selected frames/subtitles inspected; full audio transcription not verified. Visible likes/shares are not view counts.

REFERENCE ↗

我哩个豆 · Episode 41 · Northeast wife / Taipei husband

24 seconds. Domestic argument with feline characters and a physical reaction. Frames/subtitles inspected; exact voice performances not verified. Logged-out Douyin playback was possible.

REFERENCE ↗

我哩个豆 · Big Orange asking for trouble

User-supplied reference; wordplay/comedy tags. Full video analysis remains outstanding.

REFERENCE ↗

WeChat Channels reference

User-supplied reference. Not viewed successfully in this session; do not claim an analysis of its content.

REFERENCE ↗

WeChat share link

Alternate user-provided link to the same reference.

Research boundaries and evidence quality

Selected X bookmark texts/previews and TikTok favorites were sampled, not a complete watch-history audit. Tool tutorials reveal production curiosity, not necessarily the entertainment niche Eng wants to create.

Eng has an existing TikTok account with over 40K followers and prior 2021 videos. This is evidence of prior experience, not a requirement to reuse that audience.

No media pilot, voice identity, production pipeline or model-comparison result has been completed. No virality outcome is guaranteed.

Public reference links are inspiration, not proof of reuse rights. Re-dubbing and translating do not by themselves resolve permission or platform originality requirements.

07 / CAPABILITIES

Use what is here. Verify what is missing.

Tool / serviceEvidenceRole and gap
HiggsfieldAccount access verified during discoveryCandidate for image/video/voice experiments. No pilot generation or consistency test is complete; check live cost before paid work.
ChatGPT / CodexAvailable in this projectResearch, writing, planning, code, browser work and programmatic assembly. Quality still needs shot-by-shot review.
Claude Pro and Gemini Pro AISubscriptions reported by EngExact plan names and API/tool entitlements remain unverified. Eng also described his ChatGPT subscription as a “max” plan; confirm product access rather than assuming an API budget.
DeepSeek harness + Claude comparisonRequested by EngEng intends to compare this project with DeepSeek via its harness and a Claude setup he named “Opus 5.5.” Exact version availability and integration must be verified; no messages or jobs have been sent.
DescriptTools available; connection unverifiedPotential editing and transcript workflow. Verify account before relying on it.
MetricoolTools available; channels unverifiedPotential scheduling and analytics. Connect only when there is content to distribute.
Cloudflare + GitHubPublic deployment and anonymous gate checks verifiedmedia.engtaing.com serves the brief without login; studio.engtaing.com requires Cloudflare account membership across production and preview traffic. Source is in a private repository. Large-media storage is not yet provisioned.

Compare directors on the work

Proposed rubric: creative judgment 30%, ability to execute 25%, continuity and revisions 20%, factual/rights judgment 15%, cost and time 10%. Give each setup the same three reference videos and task: propose three original cat-sitcom premises, choose one, write a beat sheet, shot plan and production estimate. Compare the exact model plus its available tools; no winner has been established.

08 / THE COMMAND CENTER

One project. Many conversations.

Public front door

An intentionally shareable overview, current slate, reference links and a dated brief. HTML plus Markdown/JSON makes it portable across AI applications.

Gated working area

studio.engtaing.com contains the working desk and asset-register shell. Cloudflare account members may sign in; all workspace URLs are protected at the Worker level. No private browsing history or production assets have been uploaded.

Versioned source

One canonical project record; generated page and exports share a revision. AI chats read first, make scoped changes and return a receipt. A new chat does not automatically know this project.

Asset library

Keep footage, audio, reference rights and finished exports in private object storage or a shared drive; record IDs/links in episode manifests, not enormous files in the brief.

Later, if needed

Add a shared database or authenticated write API when concurrent editing warrants it. Start with versioned files and one publishing owner; avoid competing task boards.

The public brief is shareable. The working desk requires Cloudflare account sign-in. No private browsing history or production media is stored in this public bundle.

Open gated working desk ↗Private source repository ↗
09 / FOR THE NEXT CHAT

Pick up the work without starting over.

Share the landing-page link and ask the model to read the brief. If its browser cannot access the page or a gate, attach the Markdown file. JSON is available for tooling.

Download complete brief ↓Project data · JSON ↓

How to update the shared record

  1. Read the current revision and check the task owner/status before starting. Claim a scoped task only through an authorized shared update; a note in a separate chat is not a synchronized claim.
  2. Keep confirmed user decisions separate from recommendations. Record the source/date of material claims and mark unknowns explicitly.
  3. Return deliverables and a concise change proposal with task ID, status, evidence, decisions and next action. Do not replace the whole brief from memory.
  4. The publishing owner merges changes, increments the revision, regenerates page/exports and verifies the public/private boundary. Re-read before merging if another chat changed the same task.
  5. No automatic syncing or scheduling is implied by this page. Production spending, account changes, outreach and video publication require task-specific authority.
10 / OUTSIDE REVIEW

RV-01 · what an independent reader found.

Claude (Opus 5.5, Claude Code) · 2026-10-19 · reviewed revision 2026-10-19.2
Proposed — recorded for Eng; not yet merged into the slate, tasks or decisions

A strong handoff document with honest uncertainty labels, but heavy on process for a project with no video yet, and its pacing does not fit the 90-day reach target.

Blocking decision

MC-03, owned by Eng: CAT yes/no, per-pilot spending limit, launch date, posting cadence.

High

The 90-day target conflicts with “prove one scene first”

A fresh account reaching 100K views and 1K followers usually takes many posts, not one polished episode. The pilot, premise, character-test, episode, measure sequence can use most of 90 days before episode two. The clock also has no start: launch_date is null.

High

One small decision blocks almost everything

MC-03 (choose CAT and set a per-pilot spending limit) gates MC-05 and MC-06. Twelve tasks across four tracks plus a model comparison is more tracking than a pre-production project needs.

Medium

The source humor will not translate directly

Much of the reference comedy rests on Chinese regional dialect and stereotype (for example a Northeast wife and a Taipei husband). An English version needs its own built-in character contrast, designed from the start rather than translated.

Medium

The model comparison (MC-10) is early

It takes time from the first video and would compare models on premises for an untested format. It is more informative after episode one, using real production steps. “Opus 5.5” is a current Claude model; only plan access needs checking.

Medium

The public page carries more personal context than it needs

Some family and personal details, and the UGC host description, would be better kept in the gated workspace with a shorter public summary. Specifics are listed in the private repository review note.

Medium

Platform rules are missing from the publish checklist

TikTok and YouTube AI-generated-content labels, and YouTube’s reused or unoriginal-content monetization rules, matter most for REMIX and for any licensed or re-voiced CAT footage.

Low

Costs and rights are still blank

No track has a cost estimate yet. RED depends on adaptation rights held by others; treat it as a private craft exercise unless rights are resolved.

Proposed changes

Applies toChange
DecisionsAdd Open: launch date and cadence. Start the 90-day clock at first publish, with a minimum of 3–5 posts a week once characters pass MC-06.
DecisionsNarrow “Prove a performance before scaling” to MC-06 only; after that, volume is part of the plan.
DecisionsAdd Proposed: park RED, REMIX and UGC until CAT has published.
MC-03Clarify that it needs only four answers: CAT yes/no, per-pilot spending limit, launch date and posting cadence.
MC-04 / MC-05For each reference, name the comic engine and propose an English-native equivalent; each premise states its character contrast.
MC-07, MC-08, MC-09Status Parked until MC-11 ships.
MC-10Defer until after MC-11 and run it on real episode-one production steps.
MC-11Add AI-content labels and a reused-content check to acceptance.
Public boundaryMove family details and the UGC appearance-led wording to the gated workspace; keep a short public summary.
CapabilitiesAdd per-track cost estimates as an open gap; note RED adaptation rights.
Evidence and strengths

Evidence

Public brief.md and project.json load anonymously; the page carries noindex,nofollow (unlisted, not private).

Every studio.engtaing.com path tried (/, /api, /assets, /brief.md, /project.json, /index.html, /_worker.js) redirects anonymous visitors to Cloudflare Access.

github.com/Etaing4/media-content is private; anonymous requests return 404.

Published brief.md and project.json at revision 2026-10-19.2 were byte-identical to the repository main branch.

Not verified: preview-URL protection, because no preview hostname was available to test.

Strengths

Confirmed preferences, proposals and open questions are kept apart, and unknowns are stated rather than assumed.

The handoff rules (do not invent access, completed production, analytics or rights) suit work passed between AI chats.

Finish lines are concrete, including “Eng wants to see episode two” and the host needing a reason to return beyond appearance.