MATTSANG
01 / THE PERSON BEHIND THE PIXELS
Matthew Tsang in a white polo shirt

Hello, My Name
is Matthew,

a Southern California designer with a B.A. in Visual Communications Design from Cal Poly Pomona.

My work brings together graphic design, social media marketing, and a growing interest in AI. From product-focused visuals to experiments with creative automation, I like finding new ways to turn an idea into something useful.

PhotoshopIllustratorAfter EffectsInDesignSocial contentAI workflows
Follow the creative process ↗
02 / MY EXPERTISE

My Expertise.

Design, content, and the systems that bring them together.

01

Graphic Design

I turn product information and brand messages into clear, engaging visuals through typography, layout, and imagery.

Product graphics · Campaign visuals · Visual consistencyPHOTOSHOP / ILLUSTRATOR / INDESIGN
02

Social Media Creative

I translate campaign calendars into content for the feed, from product spotlights and customer reviews to training tips and promotional posts.

Content planning · Post design · Visual storytellingINSTAGRAM / CONTENT CALENDARS / BRANDED TEMPLATES
03

AI & Automation

I build and test workflows that connect content plans, product assets, AI image generation, and video captioning, with human review along the way.

AI-assisted visuals · Caption workflows · Connected toolsMAKE / OPENAI / GEMINI / GOOGLE SHEETS

Explore my social posts and automation case studies to see these skills in practice.

03 / SOCIAL MEDIA

Social Media Posts.

Hover for details. Click or tap to enlarge.

04 / SYSTEMS & EXPERIMENTS

AI & Automation.

Workflows, tools, and the thinking behind them.

05 / CONTACT & LINKS

Stay in touch.

SCROLL TO EXPLORE
01 / 05
POWERNET / SELECTED WORK

CASE STUDY / 01
AI & AUTOMATION / POWERNET

The 12-Image
Generator.

A Make workflow designed to turn a monthly campaign plan into draft social graphics, using visual references and real product assets.

12planned monthly posts
6reference design styles
314curated product listings
4archived test outputs
THE BRIEF

Give every draft a starting point.

Monthly campaigns need more than a headline. Each post needs a relevant product, accurate details, suitable photos, and a visual direction. The goal was to connect those inputs in one workflow rather than assemble them from scratch for every draft.

My role: creative direction, organizing the product library and content plan, choosing references, and developing and testing the automation.

01 / START WITH THE CLIENT’S CALENDAR

The direction comes from PowerNet.

PowerNet gives me an email and SMS campaign calendar for the month. It tells me what the brand is talking about, which products to focus on, and when each message is planned. My job is to use that direction as the starting point for social media posts.

For November, the calendar includes themes such as building arm strength, training with the Compact Power Bar, and organizing a game-day bag. Instead of inventing an unrelated social calendar, I translate those same themes into visuals that support the wider campaign.

Reformatted excerpt of PowerNet’s November calendar showing three campaign themes and their product direction
A readable excerpt of the supplied November email/SMS calendar. Reformatted for this case study; not a screenshot.
The starting pointClient campaign direction → a relevant idea for a social post.
02 / TURN THE CALENDAR INTO POST IDEAS

Give each image a clear job.

I turn the monthly calendar into a 12-post social content plan. Each row becomes a short creative brief: what the post should say, which product it should show, and which visual style suits the message.

For example, “Continue to Build the Arm” becomes an educational graphic showing how two ball types fit into a training routine. “What’s in the Bag” becomes a feature-led image explaining how the bag keeps equipment organized.

Archived social post plan with campaign themes translated into individual post ideas
The initial social-post plan from the case-study folder. Each campaign theme is translated into an image idea before generation begins.
Why this mattersThe AI receives a defined creative brief instead of a vague request to make a post.
03 / CREATE A CONSISTENT VISUAL STARTING POINT

Six templates give the drafts direction.

Next, I create six reusable reference templates. Each has a different purpose, while keeping the same PowerNet red, black, and white visual language. The templates establish where the headline, product image, and supporting information should sit.

I choose a template based on the post’s purpose. A product introduction needs a different layout from a customer review or a training tip. These are visual guides for the AI, with placeholders that will be replaced by the selected product and message.

The design decisionChoose the layout for the message, then give the AI a reference to follow.
04 / ORGANIZE THE PRODUCT INFORMATION

Give the workflow real products to work with.

A template alone is not enough. The system also needs the correct product photos and useful details about the equipment. I organize PowerNet’s product catalog so each brief can be connected to its matching images and descriptions.

The original collection contained 571 listings. I narrowed it to 314 products with substantial feature and benefit information. This gives the image generator more useful source material and helps keep the product shown in a draft connected to the intended brief.

Archived overview preview of the original PowerNet product catalog
Original catalog overview from the case-study folder, before the collection was narrowed from 571 to 314 products.
The ingredients for each draftA post brief + a reference template + matching product photos and facts.
05 / CONNECT EVERYTHING IN MAKE

Build a workflow that handles the handoffs.

With the content and assets organized, I connect them in Make, a tool that lets different apps pass information to one another. Google Sheets holds the post plan, Google Drive holds the reference images and finished drafts, and OpenAI creates the draft graphic.

When I start the workflow, it finds a post marked Ready that does not already have a generated image. It gathers that post’s brief, template, and product images, then sends them together to the image generator. The returned image is saved in Drive, and its link is added back to the same row in the plan.

01Pick a planned postRead its message, product, and visual direction.
02Collect the ingredientsFind its template, photos, and product details.
03Create the draftUse AI to bring the inputs together in one image.
04Save it for reviewStore the image and return its link to the plan.
Actual Make scenario screenshot showing the connected apps and steps
The actual Make workflow during its build stage. Each connected circle handles one part of finding the inputs, generating the image, or saving the result. Screenshot: October 25, before testing.
In plain languageThe workflow gathers the right material, asks the AI to create a draft, and puts the result back where I can review it.
06 / TEST, REVIEW, AND REFINE

Review the image, not just whether it ran.

I test the workflow with individual posts to check that the right brief, product, and template come together. A successful generation is one milestone; the design still needs a human review for product accuracy, readable text, layout, and how well it communicates the intended message.

These four saved draft examples show the kinds of output the workflow produces. They are AI-generated drafts, with room for design adjustments before use on social media.

Generated arm-training draft featuring plyometric and velocity balls
Arm-training education Turns the calendar’s “Build the Arm” direction into a product-focused training graphic.
Generated game-day bag draft with pocket and storage callouts
Game-day bag features Uses the “Every pocket has a job” idea to explain the equipment’s organization.
Generated sports bag draft featuring handle, wheels, and bat sleeve details
Equipment-bag spotlight Combines a hero image with a few easy-to-scan feature callouts.
Generated Compact Power Bar draft showing its different weight options
Compact Power Bar Translates the calendar’s “Small bar. Serious work.” direction into a feature-led draft.
WHAT THE PROJECT CONNECTS

From campaign direction to a draft I can refine.

The client’s calendar supplies the message. My content plan and templates supply the creative direction. The product library supplies the facts and photos. Make connects those inputs with the image generator and saves the results for review.

The goal is a repeatable way to prepare 12 monthly social graphics while keeping creative judgment in the process. These examples show the workflow and its draft output; they do not measure sales performance or time savings. Publishing is a separate step.

Creative direction meets connected workflows.

CASE STUDY / 02
AI & AUTOMATION / POWERNET

The TikTok
Auto Poster.

A Make workflow designed to turn a Google Drive video into a captioned post in a daily content queue.

1video per scheduled run
12 PMdaily check configured
AIvideo-based caption draft
1shared content queue
THE BRIEF

Make the daily handoff easier.

PowerNet has a steady stream of product and training videos. Preparing a TikTok means locating the right clip, writing a caption, supplying a video file to Buffer, a social media scheduling tool, and tracking the result.

I set out to connect those steps in one repeatable workflow. My role included planning the process, setting up the queue, connecting the services, and testing and troubleshooting the scenario.

01 / ORGANIZE THE VIDEOS

Start with one row for each post.

A Google Sheet acts as the content queue. A row holds the Google Drive video link, product name and facts, caption, approval, post date, and status. This gives the workflow a place to find its inputs and return its results.

Google Sheets module showing the content queue fields: video link, product facts, caption, approval, date, and status
Actual build screenshot: the spreadsheet fields keep each video, its caption, and publishing details together.
The starting pointA video and product details are organized together in a single row.
02 / GENERATE A CAPTION

Let the video inform the copy.

When a row is ready for a caption, Make retrieves the video from Google Drive and passes it to Gemini alongside the product context. The generated caption is returned to the corresponding spreadsheet row.

A test using a Launch F-Lite pitching-machine video produced a caption and successfully wrote it to the sheet. The result is a starting draft that can be checked for accuracy and tone.

Make caption generator showing Google Drive, Gemini upload, processing delay, and response configuration
Actual build screenshot: Make retrieves the clip, uploads it to Gemini, waits for processing, and passes the video into the caption step. This view documents the configuration in progress.
03 / PREPARE THE VIDEO FOR PUBLISHING

Connect the file to the publisher.

The publishing service needs a video file it can access. I configured a Cloudflare R2 bucket to host the file and provide a public video URL, then connected the PowerNet TikTok account to Buffer. Buffer brings the video and caption together in a post that can be reviewed and queued.

The intended workflow checks the sheet once a day at 12 p.m., processes at most one eligible video, and records the publishing response in the sheet.

Make publishing workflow with Google Sheets, Google Drive, HTTP upload, Buffer request, and row update
Actual build screenshot: the uploader connects the queue, video download, file hosting, Buffer, and the sheet update. The visible Drive error shows a troubleshooting stage, rather than a completed run.
Buffer request mapping the spreadsheet caption and hosted video URL
Actual build screenshot: the Buffer request combines the caption from the sheet with the hosted video link.
The handoffFind the next eligible video → prepare its file and caption → send it for publishing.
04 / CHECK THE BUFFER DRAFT

See the video and caption together.

The Buffer screenshot shows a real automation test draft on the PowerNet TikTok channel. The video preview and test caption are together in one place, with an “Add to Queue” option. This confirms a draft was created; it does not show a live TikTok post.

Actual PowerNet TikTok draft in Buffer with a video preview, test caption, and Add to Queue button
Actual test result in Buffer. The short caption is test copy, rather than a finished campaign caption.
05 / SET THE DAILY CHECK

Give the workflow a predictable routine.

The schedule is configured to check once a day at 12 p.m. The saved setup uses the America/New_York time zone. The intended process handles one eligible video at a time, then records the result in the sheet.

Make schedule settings showing Daily at 12 PM in America/New York
Actual saved setup view: daily at 12:00 p.m., using the time zone shown in Make.
06 / TEST AND TROUBLESHOOT

Follow the data through each step.

During troubleshooting, the Sheets search returned no rows because its exact status filter found no matching entry. A separate test ran Gemini alone without the video data from the earlier modules. These checks showed why the status in the sheet and the full scenario run both matter.

I checked the row selection, video link mapping, and caption output in stages. This approach helps identify which handoff needs attention without assuming that a successful module means the whole workflow has finished.

WHAT THE PROJECT CONNECTS

A foundation for a repeatable content process.

The saved work includes a video-based caption test, a hosted video upload, and a test draft in Buffer. The daily schedule and one-video limit were configured. Together, these show how the separate parts support a repeatable publishing process.

A full scheduled run from a new row through to a live TikTok post has not been verified in the available records. Time savings, sales impact, and total costs have not been measured.

Creative direction meets connected workflows.