A reference for the next scene
Inspect each persona’s reference candidates and recent work, then choose which reference to use.
Case 06 · Persona Studio
Identity, images, video.
One production workspace.
I integrated AI tools into a local product for defining persona references, creating images and videos, and reviewing the results. Creators can describe their intent in Korean and inspect the output before deciding what to use.1
Local Mac prototypeIn development
- My role
- Problem definition · product design · AI implementation direction · integration and validation
- Problem
- Creative context and review become fragmented across separate tools
- Unit of production
- Persona workspaceReferences and production history together
- Connected media
- Image → VideoOutputs connected in one production experience
- Selection of results
- Human reviewCompletion and quality judgment remain distinct
Product experience
Keep the context that repeat production needs.
Persona references, creative intent, and completed outputs share one workspace.
Write in Korean. Check the wording.
Describe scenes and motion in Korean, then inspect and edit the English generation text or discard changes.
A history of images and videos
Enlarge images, play videos, and see approval or hold decisions alongside the output files.
Design judgment
Separate automation from creative judgment.
The production experience makes the creator’s decisions explicit, even where tools handle the work.
Who defines a consistent identity?
Repeat generation alone cannot establish whether a character still reads as the same person. I made reference selection an explicit user decision, clarifying the boundary between automated work and creative responsibility.
A valid file is only one measure
A successfully created file does not establish the quality of an expression, hand, or movement. I treated functional validation and visual assessment of actual outputs as separate checks.
Intent must survive translation
Automatic translation can change the intended meaning. I kept the result visible and editable so the creator retains control over the wording used for generation.
Evidence & statusValidate the workflow.
Validate the workflow.
Judge the actual output.
Validation covers both the connected image-to-video workflow and the creator’s review experience. The scope below comes from saved validation records.
Current product and validation scope
A local Mac prototype connects reference selection, image and video creation, and approval or hold for three personas. Tests verify application behavior and output files; people assess identity consistency and visual quality from the results.
- Saved test result442
- A saved run reports 44 passing tests covering application behavior and output-file validity. Functional validation is separate from visual-quality review.
- Image-to-video integrationCompleted
- Saved checks confirm full decoding and browser playback of a generated video.
- Authoring and reference selectionUI · API
- UI and API checks cover manual translation edits, cancellation, and the selection or hold of reference candidates.
My contributionCreative judgment,
Creative judgment,
built into the product.
I decided what to automate and where the creator should remain in control. I designed the product flow, directed AI implementation, and reviewed and integrated the results into one production workspace.
- Product problem definition
- Workflow design
- AI implementation direction
- Integration and quality judgment
Sources and scope
- Prototype guidance and current UI/source inspection support persona-reference management, Korean authoring with English review and editing, image/video creation, and review. The role describes user-led problem definition, design, and AI implementation direction.
- Saved functional-validation records report 44 passing tests, full video decoding and browser playback, and authoring/reference-candidate review checks. Visual quality and identity consistency are assessed separately.
Detailed production methods and operational design are not disclosed.