Creator Flow: Complete Workflow and System Behavior

Creator Flow is Cannon Studio’s production pipeline for long-form consistency. Instead of generating disconnected clips, it builds a structured project with reusable world context, visual anchors, and per-shot controls.

At a practical level, Creator Flow is best understood as a stack:

• Project and universe context
• Chapter and scene narrative structure
• Character and location grounding assets
• Shot-level image, motion, narration, and audio controls
• Scene/chapter stitching and final delivery

Learning goal for this guide:

By the end, you should be able to produce more consistent outputs on the first pass, know when to regenerate versus refine, and understand why some prompt choices perform better than others in multi-shot storytelling.

Core mindset:

• Treat each shot as part of a sequence, not a standalone image
• Prioritize continuity decisions before polish decisions
• Use prompts as production instructions, not descriptive paragraphs

You can start from scratch or start from an existing universe. The right choice depends on whether your priority is speed or continuity.

• Scratch: use when building a new world, visual style, and story direction.
• From Universe: use when extending an established world and preserving continuity.

For best outcomes, spend extra care on three setup fields:

• Story premise and constraints
• Character/object roster expected to recur
• Key locations with visual identity language

Best-practice setup prompting:

• State genre, tone, and pacing in one concise paragraph
• Define visual style with 3-5 repeatable anchors (lighting, texture, palette, lens feel, realism level)
• Name recurring characters/objects explicitly so they can be grounded later
• Describe locations in terms of repeatable visual identity, not one-off scene details

Common setup failure:

Overwriting too much context when extending a universe. When using an existing universe, focus on what is new and additive rather than re-defining the whole world.

Creator Flow intentionally separates story hierarchy so you can iterate at the smallest useful scope.

• Chapter: macro narrative beat and progression objective.
• Scene: local unit of story action in one setting/time context.
• Shot: executable asset unit with explicit generation settings.

This hierarchy is why you can rework one shot without redoing the entire scene, and one scene without reauthoring an entire chapter.

How to write strong chapter/scene intent:

• Chapter: what must change emotionally or narratively by the end
• Scene: what conflict or reveal happens in this location/time context
• Shot: what the audience should notice first, then what should change

Fast quality test:

If you can summarize each scene in one clear sentence without mentioning camera terms, your narrative layer is strong. Camera language belongs at shot level, not scene purpose level.

Character definitions are grounding records, not just labels. They are designed to reduce identity drift across multiple generated assets.

• Primary identity fields shape recurring appearance and behavior
• Variants represent controlled alternate looks or states
• Object-type characters are useful for recurring non-human entities

Effective pattern: keep one canonical base character and create only a small set of purpose-specific variants. Too many loose variants reduce consistency and increase selection overhead in production.

Character prompting best practices:

• Describe immutable identity markers first (face structure, silhouette, key wardrobe signatures)
• Then describe variable details (mood, temporary styling, scene-specific props)
• Avoid contradictory terms like “rugged and pristine” unless intentionally stylized
• Use variants for deliberate changes, not random experimentation

When to create a variant:

• Costume/state changes that recur across multiple shots
• Battle-damage or weather-state continuity
• Distinct role states (civilian vs tactical gear, day look vs night look)

Locations are the continuity backbone for environment consistency. Zones and angles let you preserve geography while changing composition intentionally.

• Location: master environment identity
• Zone (sublocation): reusable sub-area within a location
• Angle: repeatable framing viewpoint for predictable shot continuity

If your scene repeatedly returns to a place, formalize it as a zone with named angles. This removes ambiguity and improves reference selection quality during generation.

Location prompting best practices:

• Define environment DNA first: materials, architecture language, weather profile, ambient mood
• Then define cinematic qualities: lens feel, depth behavior, contrast profile
• Use stable naming for zones and angles so reuse is obvious across scenes

Angle strategy:

Name angles by purpose, not only direction. For example, “Interrogation Close 3/4 Left” is more useful than “Left Angle 2” because it encodes shot function and composition intent.

Pre-production is where you lock the opening visual frame and shot intent. High quality output usually comes from strong pre-production discipline.

• Write the image prompt as an opening-frame directive, not a full motion description
• Validate selected character/location grounding before generation
• Use prompt enhancement tools to tighten clarity, not to add unnecessary complexity

Recommended image prompt structure:

• Frame type and subject priority
• Composition and relative placement
• Lighting and atmosphere
• Environment detail and realism constraints
• What should not be present (when necessary)

Before you click Generate:

• Check that the shot location is correct (locationId and optional sublocationId)
• Check that selected character identities match this scene beat
• Check that your opening frame matches narrative intent, not just “looks cool”

Shot Production converts the still start frame into motion output using duration, quality, and grounding references. Keep motion instructions concise and physically plausible.

• Use camera verbs and action verbs that are easy to execute
• Use explicit references only when continuity risk is high
• Match quality and duration to narrative purpose and budget

Tip: if your shot has character turns, entrances, occlusion, or identity-sensitive action, grounding references are usually worth the extra setup.

Recommended video prompt structure:

• Action beat: what changes from frame 0
• Camera behavior: lock-off, push, pan, tilt, track, handheld feel
• Pacing: subtle, controlled, sudden, aggressive
• Continuity clause: what must remain consistent (identity, setting, lighting logic)

Duration and quality guidance:

• Use shorter duration for single-purpose visual beats
• Use longer duration only when motion arc needs time to read clearly
• Higher quality is most valuable on hero shots and close facial storytelling

Post-production is where timing polish and perceived quality are often won. Use narration timing, sound levels, and stitch previews to verify rhythm.

• Normalize narration and SFX levels before final stitch passes
• Validate lip-sync and narration start offsets shot by shot
• Use scene/chapter preview durations to catch pacing drift early

Post workflow that saves time:

1. Finalize visual timing first
2. Then set narration timing and tone
3. Then set SFX and music balance
4. Finally stitch and review sequence rhythm

What to listen for:

• Narration clarity during high-energy moments
• SFX peaks that mask dialogue
• Abrupt audio texture changes at shot boundaries

Autopilot is ideal for broad first-pass generation. Bulk generation is ideal when you have already reviewed upstream assets and want fast, parallel execution.

• Use Autopilot for exploratory iteration and production acceleration
• Use bulk video generation after images are approved
• Always check total credit cost before running large parallel waves

Recommended production strategy:

• Phase 1: generate broad image coverage fast
• Phase 2: review and fix hero shots
• Phase 3: run bulk video generation on approved image set
• Phase 4: do targeted polish instead of broad re-generation

This pattern usually gives the best speed-to-quality ratio while reducing wasted regeneration spend.

Creator Flow does more than pass your prompt directly into generation. It helps shape the final prompt based on scene context, selected references, and continuity needs.

• Image prompts are treated as opening-frame instructions, not full motion scripts
• Video prompts focus on movement, camera behavior, and continuity from that opening frame
• Character references are emphasized when identity drift risk is high (turns, occlusion, entrances)
• Location and zone references are emphasized when environment continuity matters across shots
• If explicit references are unnecessary, prompts stay lighter to avoid over-constraining output

In practice, this means generate-shots tries to balance two goals: preserving continuity and keeping each shot flexible enough to match your intended action and framing.

When references are most valuable:

• Character profile turns or partial visibility
• Complex entrances/exits where identity can drift
• Environment-sensitive moves where background continuity must hold
• Shots where start frame and end intent need tighter alignment

When lighter prompts usually perform better:

• Simple single-subject beats with clear composition
• Shots where heavy constraints can fight desired motion
• Early exploration passes before final continuity lock

A good rule: constrain only what must stay stable, and leave the rest available for natural motion and composition adaptation.

• Characters: consistent face, silhouette, wardrobe, and identity markers
• Locations: lighting/weather continuity and zone correctness
• Shots: opening frame matches intent before motion starts
• Audio: narration volume and SFX levels are balanced for dialogue clarity
• Stitch preview: chapter-level pacing feels intentional, not rushed

Use this review order:

1. Continuity errors (identity, location, angle)
2. Narrative clarity (does each shot communicate its purpose?)
3. Cinematic quality (composition, lighting, motion smoothness)
4. Audio intelligibility and emotional rhythm