# Studio Systems / R01

Legend × Quik · Immortal Legend Studios × QuikyCode · 16 October 2026.

A shared production plan for both creators and their agents. Choose the work, choose where it runs, preserve the evidence, and hand off a useful result. The tool register is a dated snapshot; each account and machine needs its own connection check.

## Cloud and local: the team plan

| Lane | Useful work | What must be ready |
|---|---|---|
| Cloud / start now | Search accessible project sources, recover prompts, compare asset lists, plan scenes, draft code and prepare reviews. Either creator can direct this from a browser. | The selected agent needs the actual sources and its own service permissions. No local GPU is needed for cloud work. |
| Cloud / add for a job | Repeatable image workflows, remote rendering, builds and longer jobs. | A configured service, a successful sample, usage limits and an output destination. No Comfy or render farm connection is established by this guide. |
| Local / each workstation | Blender modeling and camera renders; Unreal asset import, collision, gameplay and performance checks; suitable local-model experiments. | A stable machine, installed tools, project access and a small execution test. Legend's PC1 is under diagnosis; Quik's readiness is not established here. |
| Shared / every handoff | Source revisions, task ownership, artifacts, checks and decisions. | One repo record and linked outputs. Private agent memory is not the team database. |

The work laptop can direct cloud tasks and review results while PC1 is unavailable. It does not need Blender, Unreal or a local LLM installed for that role. Software installation and computer control apply to the explicitly connected target; access here does not provide access to either creator's desktop.

## How Legend and Quik work together

Both creators can write, design, model, code, test and review. The useful starting responsibilities are Legend for creative intent, canon and visual acceptance, and Quik for engine integration evidence. These are shared working responsibilities, not exclusive skill boundaries.

- Give each task one accountable owner and name the reviewer. Each person's dot can coordinate their assigned work.
- A builder produces one bounded output on its own branch or in its own artifact folder. A reviewer checks the result against the source and acceptance criteria.
- Share task briefs, commits, images, logs and review notes in the repo/Production Desk. Direct dot-to-dot messaging and automatic shared memory have not been established.
- Coordinate Unreal binary-file ownership and LFS locks. Two agents should not edit the same map or asset concurrently.
- Label results DRAFT, REVIEW, ACCEPTED or BLOCKED, and record who accepted them. A successful script is not proof that an Unreal asset or world works.

## Subagents and automation

Subagents are useful for independent, scoped work: an archive researcher, a prompt curator, a builder, and a separate reviewer. Parallel work helps when each job has clear inputs and its own output. Shared files and one coordinator keep their results aligned.

A coding/testing loop should stop at a defined result, retry count, time limit or blocker. Do not describe session subagents as an endlessly running studio. Unattended work needs a configured runner or scheduler, persistent task state, usage limits and a tested completion path.

Useful future automations include a weekly source-change digest, missing-preview reports and asset-register checks. They are proposals; this release creates no schedule or background job. Service/API subscriptions and local compute costs depend on the chosen route and need a per-job limit.

## Scenes, old prompts and new storyboards

Yes: either creator's agent can find accessible scene text, search the prompt archive, match existing art and plan new images. Begin with the repo's `story/` and `art/midjourney/` records; they are a starting archive, not proof of a complete Midjourney account export.

1. Identify the actual scene, edition, beat, era, location, action and characters. Check current decisions before older chapter drafts; novel and game versions keep separate records.
2. Recover the original prompt text and image, plus job ID/URL, model/version, parameters and reference images when available. Mark missing fields unknown.
3. Match candidates by face, wardrobe, location, mood and lighting. Record why each matches and whether it is approved, historical or a candidate.
4. Plan the minimum useful shot sequence: establishing view, action coverage, reaction and story detail. Keep screen direction, blocking, wardrobe and environment continuity.
5. Generate missing frames using the available image tool or a connected provider. Adapt old prompts to the selected model. A prompt/seed does not guarantee the same image across models or versions.
6. Deliver a contact sheet, individual frames, shot list, prompts, source IDs and continuity notes. Both creators review; creative and engine acceptance remain separate.

Midjourney supports personal-history searches and copying job/prompt information. Use an authorized export/intake route for account material. Its published rules prohibit unauthorized automation; do not presume a general API or scrape the account. An image download is not proof that its original prompt was included. When only pixels remain, label any inferred prompt **reconstructed**.

Proposed first pilot: one scene, six storyboard frames and two polished images. Legend brings scene intent and references; Quik adds useful environment, prop and camera coverage. Review the contact sheet before expanding. No images are commissioned by publication of this plan.

## Spatial creation: another route to consistent images

Use the adopted [Spatial Imagery workflow](/protocols/spatial-imagery/) for views that must describe the same physical place. Establish a reviewed layout, shared 3D model and saved cameras; render those views, optionally finish them with controlled image generation, then compare against the geometry.

For Crown Key, a waterfront view must use the actual reviewed building placement and camera direction to establish what appears beyond the windows. A visually convincing generated view alone cannot establish the building's distance from water or its metric scale.

Freeform imagery remains useful for exploration, style and mood. Geometry supplies spatial constraints; image finishing supplies appearance. Record the source, layout/model/camera revisions, method, differences and approval scope. Comfy workflows are a candidate way to automate a repeatable finishing process after a sample connection succeeds.

## First shared work queue

These are suggested tasks, not running jobs or construction authorization.

| Priority | Deliverable | Suggested ownership and acceptance |
|---|---|---|
| 1 / Inventory | Reconcile Quik's 105-item worklist against available source files and engine evidence. | Quik supplies current asset evidence; Legend and an agent organize the register. Verify each source/status instead of inventing completion. |
| 2 / Visual pilot | One scene packet, prompt/source register and reviewed six-shot storyboard. | Either creator owns the task; the other reviews continuity and usefulness. Preserve original versus reconstructed prompts. |
| 3 / Foundation prep | Crown Key / southwest West Coast source and connection packet with unresolved measurements named. | Legend supplies visual direction; Quik identifies engine/frame requirements. An agent prepares the comparison and handoff. Fitted construction remains paused. |

Current Building and District revisions come from [the protocol pointer](/protocols/current.json). Use the current [Scale Register](/production/scale/) and [shared-world brief](/protocols/world/). The freestyled Grove geography remains archival; no final island size, master map or launcher readiness is approved by this page.

## Start either creator's next task

Download the task template and fill in a task ID, owner, reviewer, goal, exact source revisions, execution target, dependencies, allowed changes, output paths, acceptance checks and stop limits. For creative work include canon/visual status; for code include branch and tests; for engine work include project/world and binary-file ownership.

Start with one small test for each new connection: read a named source, produce a disposable output, retrieve it and record the evidence. Move a tool from planned to verified only after that test succeeds. Each creator connects their own account and workstation as needed.

## Maintaining this area

Studio Systems is the umbrella for dots, subagents, plugins, computer control, creative pipelines, local models and future services. Add tools to the versioned register with a category, purpose, execution location, status evidence, owner/setup requirement and next useful test.

Edit the repo sources and run `python tools/build_production_desk.py`. Keep generated source/deployed pages together; update this release and its date when capabilities change. The website displays a snapshot, not live agent telemetry or a shared task backend. Secrets and unrelated private content stay outside the publication inputs.

## Official capability references

- [Dots: computers and apps](https://learn.chatgpt.com/docs/dots/computers-and-apps)
- [Comfy: agent connections and workflows](https://docs.comfy.org/agent-tools/mcp)
- [Midjourney: search and organize creations](https://docs.midjourney.com/hc/en-us/articles/33329462451469-Organizing-Your-Creations)
- [Midjourney: job and prompt details](https://docs.midjourney.com/hc/en-us/articles/33390732264589-Creating-on-Web)
- [Midjourney: automation rules](https://docs.midjourney.com/hc/en-us/articles/32013696484109-Community-Guidelines)

References checked 16 October 2026. The register's connection evidence is separate from a vendor's advertised capabilities.
