PRISCILLA TSE NOK IN — AI PRODUCTION COORDINATOR

AI production
for what’s next.

I am an AI production coordinator with four years of experience at Blink Production HK. I manage AI image and video projects from the initial brief and visual exploration through feedback, refinement and delivery. My toolkit includes Midjourney, Seedance, Kling, ComfyUI, Claude, Gemini, Magnific and DeepSeek, selected according to each project’s creative and production needs. I am especially interested in the intersection of generative media, digital culture and Web3. I understand how emerging tools become clear, useful content—and how strong coordination turns ambitious ideas into work that teams can confidently support, approve and share within a practical, commercial production setting. I have also been an active digital-asset market participant since 2022, and I am now looking to bring both sides—applied AI and market literacy—into a Web3 team.

↓

NDA-SAFE SELECTED WORK

03 CASE STUDIES

INTERVIEW EDIT

(01)

HYBRID PRODUCTION / PHOTOGRAPHY + AI

From plain-screen capture to social-ready worlds

✳
Plain-screen source photograph of a chocolate spread jar
01 / BASE CAPTUREControlled studio source
AI-assisted styled social image of the chocolate spread jar
01 / AI OUTPUTGenerated environment and art direction
Plain-screen source photograph of a coffee cashew jar
02 / BASE CAPTUREControlled studio source
AI-assisted styled social image of the coffee cashew jar
02 / AI OUTPUTGenerated environment and art direction

PROJECT CONTEXT

I developed a hybrid product-image workflow designed to turn a simple studio capture into a broader library of social-ready product visuals. The product was photographed against a plain screen so I could begin with a clean, reliable reference while keeping the packaging, proportions and materials grounded in a real object.

TOOLS USED

Controlled studio photography; retouching; ComfyUI-based image workflows; multiple generative image models for image-to-image exploration; Magnific for detail enhancement and final-resolution output.

APPROACH

I first created clean base photographs with neutral lighting and minimal background information. I then used the product captures as structural references across several AI image models, testing different ways to build settings, props, lighting and mood around the original pack shot. I compared each model for product fidelity, realism and art-direction control, then refined the strongest outputs through targeted iterations and finishing. The aim was not to replace photography, but to combine the accuracy of a real shoot with the speed and creative range of generative production.

OUTCOME

The workflow produced a flexible series of styled product images from a small set of controlled source photographs. It demonstrates how I can extend one shoot into multiple content directions for social media while maintaining a recognisable product. The images shown pair the plain-screen captures with the finished AI-assisted scenes.

(02)

END-TO-END STORYTELLING / STORYBOARD + VIDEO

One-person AI production, from character to final sound

✳
FINAL VIDEO / 01:20Concept, character design, storyboard, image and video generation, audio, SFX and editing by me
Minions-inspired character design for Daniel
CHARACTER 01 / DANIELBespoke design based on a real participant
Minions-inspired character design for Borna
CHARACTER 02 / BORNABespoke design based on a real participant
Minions-inspired character design for Tesleem
CHARACTER 03 / TESLEEMBespoke design based on a real participant
Minions-inspired character design for Garth holding a guitar
CHARACTER 04 / GARTHBespoke design based on a real participant
Thirteen-scene storyboard with visual action, dialogue, sound and camera notes
STORYBOARD / PRODUCTION MAP13-scene production storyboard covering action, dialogue, sound, timing and camera direction

PROJECT CONTEXT

I developed a character-led video as an independently managed AI production. Four bespoke characters were designed from the visual traits and personalities of real people, translated into a playful Minions-inspired world. I used the storyboard as the production backbone, defining character roles, comic timing, camera movement, dialogue, music cues and sound effects before generating the final scenes.

TOOLS USED

Claude, Gemini and DeepSeek for concept development and planning; Midjourney and ComfyUI for character and image development; Seedance and Kling for motion; AI audio tools for voice, music and SFX; editing tools for assembly and finishing.

APPROACH

I owned every stage of the pipeline: story development, character design, shot structure, storyboard, image generation, video generation, audio, SFX and final editing. I broke the narrative into numbered scenes with visual action, dialogue, sound and production notes, then used that structure to monitor progress, identify missing shots and control continuity. I tested outputs against character consistency, performance and pacing, revised weak scenes and kept the project moving without external task management.

OUTCOME

The result is an approximately 80-second film that I produced and edited independently from concept to final sound. The finished demo, character studies and storyboard show how I turn an open-ended idea into a production-ready sequence, manage milestones and carry a multi-stage AI project through delivery. Before public sharing, I would confirm participant consent and third-party style or music clearance.

(03)

INFORMATION DESIGN / ART DIRECTION + AI

AI infographics, independently art-directed

✳
Chinese-language seatbelt safety infographic in yellow and navy
INFOGRAPHIC / PUBLIC SAFETYSingle-message hierarchy and symbolic art direction
Chinese-language Hong Kong travel offers infographic
INFOGRAPHIC / TRAVEL GUIDEModular information design for quick comparison

PROJECT CONTEXT

I created these infographics as independently directed, AI-driven visual communication projects. Each began with a practical information challenge—making public-safety guidance or a dense list of travel offers quick to scan, easy to remember and visually suited to its intended audience.

TOOLS USED

Claude, Gemini and DeepSeek for research support and content structuring; generative image models for illustration and visual development; AI-assisted typography exploration; layout and image-editing tools for hierarchy, checking and production-ready finishing.

APPROACH

I handled the full process myself: research, information hierarchy, copy structure, art direction, visual system, image generation, layout and final checks. I first identified the one message the audience needed to retain, then organised supporting details into a clear reading order. I used AI to accelerate visual exploration while actively reviewing text accuracy, icon meaning, contrast, legibility and consistency. I also set my own milestones, monitored open tasks and revised the work until the information and visual tone supported each other.

OUTCOME

The finished pieces translate different kinds of information into distinct but accessible visual systems: a bold safety message built around one memorable symbol, and a modular travel guide organised for comparison. Together, they demonstrate end-to-end ownership, audience awareness and the discipline to manage AI-generated content as a communication product rather than a one-off image.

SHIPPED PRODUCTS

Building,
not just briefing.

Beyond client production, I build and ship my own products using AI-assisted development—Claude for architecture and design, Codex for implementation, and Gemini API for inference. Both are deployed and in daily use.

(04)

CAL LO MI / AI NUTRITION TRACKER

Making AI useful
in daily life.

A meal-logging and nutrition-tracking app for two users. Photo, album or text input is processed for calorie and macro estimation, then confirmed by the user before it commits—a human-in-the-loop pattern that keeps AI convenience without inheriting AI error. Personalised targets are calculated from biometrics, including TDEE, NEAT and daily macros, with Apple Health and Garmin activity integration.

The data model distinguishes AI-estimated values from verified database values, so the user always knows what they are looking at. Paired users share an activity feed, comments, challenges and a shared inventory with expiry tracking and recipe suggestions. Account-level separation keeps personal data private while syncing shared data to the cloud.

STACK / CLAUDE · CODEX · GEMINI API
Currently exploring local LLM hosting to reduce inference cost.

CAL LO MI daily nutrition dashboardCAL LO MI AI meal input optionsCAL LO MI paired-user activity feedCAL LO MI Cantonese AI coach
(05)

TONIGHT WE PICK / DECISION-WEB APP

One problem.
One interaction.

VIEW LIVE ↗

A decision tool for couples facing daily choice fatigue. Deliberately narrow: one problem, one interaction, no feature creep.

LIVE / BLACKMI-PP.GITHUB.IO/TONIGHT-WE-PICK

Tonight We Pick home screen with decision categoriesTonight We Pick paired-user setup screenTonight We Pick random decision wheel

WHAT THESE TAUGHT ME

The distance between an idea and a working product has collapsed. What used to need a technical co-founder now needs product judgment and fluency with AI-native tooling. That shift is the same one reshaping how teams inside AI-first companies operate—and it is why I want to work inside one.

AI ECOSYSTEM OBSERVATIONS

What four years
in the trenches
has taught me.

Working across Midjourney, Seedance, Kling, ComfyUI, Claude, Gemini, Magnific and DeepSeek every week has given me a specific kind of literacy: not just how to use each tool, but how they compete, where they overlap, and what commercial workflows they actually enable.

01

Tool landscape.

The image-generation space has moved from single-model dominance to specialised stacks—Midjourney for stylised concept, Gemini for prompt adherence, Magnific for high-fidelity finish, and ComfyUI for granular control. Video is now competitive across Seedance and Kling for short-form. The Chinese and Western tool ecosystems are diverging in interesting ways—a dynamic under-appreciated in Western coverage.

02

Commercial pattern.

The projects that succeed commercially are the ones where AI extends existing production capacity rather than replacing it—hybrid workflows, not full automation. Clients still value human judgment on brand fit, taste and final approval.

03

Where AI meets Web3.

The convergence I find most interesting: on-chain provenance for AI-generated assets, generative content pipelines for gaming and digital collectibles, and AI-assisted worldbuilding for open-metaverse experiences. This is where I think Animoca’s positioning across gaming, tokenisation and AI investment creates real partnership opportunity—and where I would want to contribute.

WHERE THIS GOES NEXT

Creative fluency.
Commercial focus.

AI production taught me how to connect emerging technology to outcomes people can actually use. I translate between creative, technical and commercial stakeholders, turn open-ended ideas into workable plans, and hold momentum through testing, feedback and delivery.

In business development—

Partnerships depend on clear communication, credible execution and knowing the space well enough to be useful in the first conversation. I have built a client book from zero under financial-services regulation, and I have been an active digital-asset market participant since 2022, following institutional adoption closely: MicroStrategy’s corporate treasury model and its structured exposure instruments, Web2 operators migrating on-chain, and Hong Kong’s licensed exchange and stablecoin build-out.

In AI product—

I have coordinated commercial AI production for four years and shipped two products of my own. I know where generative models are reliable and where they are not, what human-in-the-loop actually costs to design, and how to keep an experimental project moving toward a shippable outcome.