Skip to main content

AI-Assisted Cinematic Production

Rum & Velvet — The Island Spell

An AI-assisted cinematic music film built to hold character identity and continuity across a complete production, with governed assets and frame-by-frame quality control.

Status

Delivered to competition

Role

Director, producer, character and identity designer, prompt and shot author, editor, and quality controller.

Current milestone

The film is complete and was delivered to competition. The production structure, anchors, continuity QA records and final masters are retained.

Technology areas

Image Generation, Motion Generation, Identity Lock and Character Anchoring, Continuity QA

The problem

What the project is designed to address.

AI-generated film fails visibly at continuity. Faces drift between shots, wardrobe changes without reason, and no one can say which version of an asset was approved. The individual frames are convincing; the sequence is not.

The solution direction

A structured response to the operational need.

A production pipeline built around identity anchors and explicit continuity review: character references generated and locked first, every shot checked against them, and every asset versioned with its provenance recorded.

Project context 01

What this proves and what it does not

It demonstrates that identity and continuity can be held across a complete AI-assisted production, with governed assets and a reviewable quality process.

It does not demonstrate commercial performance. The film was delivered to a competition; no audience figures, placement or revenue are claimed, and none should be inferred.

Target users

Who the project is intended to support.

01

Competition judges assessing AI-assisted film

02

Producers evaluating whether AI production can hold a narrative

03

Practitioners who have hit the identity-drift problem themselves

Project objectives

What the initiative is intended to accomplish.

  • Hold a single recognisable character identity across a whole production
  • Make continuity a reviewable stage rather than a hope
  • Keep asset provenance traceable from generation to final master
  • Deliver a complete piece to a competition deadline

Mohamed Sheriff’s responsibilities

Founder-led product and solution development.

  • Character design and identity-lock approach
  • Shot authoring and cinematography decisions
  • Continuity review and frame-by-frame quality control
  • Editing and final delivery

Architecture and workflow

How the system is put together.

  • 01

    Character anchors generated and locked before any scene work began

  • 02

    Shot-by-shot generation checked against the anchors rather than against the previous shot

  • 03

    Asset governance with staged directories: character anchors, environment anchors, scene heroes, props and inserts

  • 04

    Continuity QA as a distinct stage with its own pass criteria

  • 05

    Final-delivery masters kept separate from working assets

Work completed

Documented deliverables and implementation milestones.

Completed and verified. Planned work is listed separately under the next phase, and the two are never mixed.

01

A complete cinematic music film delivered to competition

02

Character anchor set with an identity-lock approach applied across shots

03

Governed asset structure separating anchors, scene heroes, props, staging and final masters

04

Continuity QA pass over the assembled sequence

05

Production bible documenting the approach

Key decisions

Important choices shaping the project.

01

Anchors first, scenes second

Generating scenes and hoping identity holds is the standard failure. Locking character references before any scene work made drift detectable rather than inevitable.

02

Check each shot against the anchor, not the previous shot

Comparing to the previous shot lets identity drift compound invisibly, one acceptable step at a time.

03

Continuity QA is a stage with its own criteria

Folded into editing, continuity gets traded away under deadline. As a separate gate it has to be passed.

Constraints

What remains unresolved or incomplete.

  • Identity holds well but is not perfect; some shots required regeneration and a small number of compromises remain
  • The approach is labour-intensive and does not yet scale to longer-form work without more automation
  • Continuity review is manual, which is why it is reliable and also why it is slow
  • No audience or performance metrics are claimed; the outcome recorded here is delivery to competition, nothing more

Next steps

Where the project moves next.

  1. 01

    Automate parts of the continuity check that are currently manual

  2. 02

    Generalise the anchor and governance model into the Automated AI Media Factory

  3. 03

    Extend the approach to longer-form work, where the labour cost currently becomes prohibitive

Related consulting capability

AI Media and Digital Identity Studio

The identity specification, reference assets, production workflow and asset governance built here are exactly what the AI Media and Digital Identity Studio engagement delivers for a client.