Your agent — Claude or ChatGPT Codex — does 90% of the production: the reference sheets, the script, the shots, the voice, the music, the edit, the posting. You do the ten percent that actually matters. This is the class where you stop clicking and start directing.
The hard part of AI video was never the first shot.
It was the hundredth — with the same face.
Consistency is the wall every AI creator hits. Cross it, and you don't have clips any more. You have a production.
Level is what you can make. Format is how you learn it. Level 2 is taught face-to-face only — a pipeline this involved needs someone at your shoulder.
Straight from the booking system, so what you see is what is open. Choose a date and it comes with you — you set up your account on the way, and that date is waiting for you when you land. Every date is face-to-face at Paya Lebar Square (Office Lobby 1): a comfortable room, quick Wi-Fi, coffee and tea on tap, and a short walk from the MRT. All times Singapore.
Three and a half hours in the room with Dr. Andrew Chew, hands-on throughout, with follow-up support after class — and 10% back in points towards your next class. Your agent and Higgsfield subscriptions are separate — see what you need. Face-to-face classes need a minimum number of participants to go ahead; if a date doesn't reach it we may ask you to move to another, and everything you've paid carries over.
That's the whole difference. In Level 1 you direct every shot yourself. Here, an agent runs the pipeline and you approve the output. Same craft, different amount of hand-work.
Cost to produce is not the course fee. It's what the tools charge you to generate each finished video — Level 2 leans on heavier generation, so it costs more per film. Course fees are the same for both levels.
No inflated numbers, no rented audience — the account is brand new. What matters is that these are real posts on real platforms, and every frame of them came out of the pipeline you're about to learn. A comedy, a fight scene, a fragrance campaign — same character, same studio. Judge the work, not the follower count.
Justme Jasmin is an AI character created by Dr. Andrew Chew, and is openly labelled as AI on every platform.
This single artefact is what separates a lucky render from a repeatable production. Build it once, and every future shot has something to obey. You will build yours in Hour 1.
Not the prompting. Not one step. The production. From the reference sheets that hold your character together, to the script, to every clip, to the voice, the music, the edit, the captions — and out the door to the platforms. Read that list again slowly. That is the whole job.
Your agent builds the character sheet — five angles, one face — and the scene sheets that lock the world she lives in. Every asset gets logged and reused, so shot forty still looks like shot one. This is where consistency stops being luck and becomes a filing system.
It scores your idea out of twenty and tells you honestly when it's a twelve — then offers three better ones. It writes the beats, cuts every line to fifteen words, and builds the locked-world table that stops the model redecorating your kitchen between shots.
Stills first — cheap to iterate, and you choose the image model every production: Nano Banana 2 (cheapest, fastest), GPT Image, or Seedream 5.0 Pro (strongest at following instructions). Switch mid-reel if one blocks a prompt — you are always the final judge. Then each approved still becomes the first frame of its clip, generated on Seedance 2.0 — the newest and most powerful video model available today. Native audio, real performance, and the fidelity that makes people forget to ask whether it's AI. That first-frame anchor is precisely what stops your character morphing mid-reel. It fires the generations in parallel, polls them, retries the failures, and files everything where it belongs.
A cloned voice — hers, or yours — carrying the accent and the cadence, because get the voice wrong and the comments will roast you. Then an original Suno track scored to the mood of the cut. No library, no licensing, no copyright headache.
Crossfades, audio blends, burned-in captions positioned above the Instagram UI zone, exported to spec. You never open an editor. If the caption timing is off, you say so, and it re-renders in seconds.
Out to the platforms, AI-labelled — which is required, and also the best content angle you have. Forty-eight hours later it logs the views, the shares and the comments people actually quoted. Those quotes are next week's content.
So what's your ten percent? Taste.
Deciding who she is. Knowing which take is the one. Saying no at the gates. The machine will happily produce a hundred beautiful, forgettable reels — judgement is the part it can't do for you, and it's the part this class actually teaches.
You leave this class with a folder of files. Drop them anywhere, point your agent at them, and it reads the runbook, scaffolds your entire studio, and walks you through five production phases to a finished reel. That folder is the course.
Automation without a handbrake is just a faster way to make rubbish. Nothing expensive gets generated until you've said yes.
Not a frame is generated until you approve the storyboard. Ideas are free. Video is not.
Every frame checked before it moves: does the face match the sheet? The room? Is the prop the right size?
Face, scene, realism, acting. All four, or it goes back. You don't ship "close enough" at 1am.
Face-to-face with Dr. Andrew Chew in a small room. Hands-on from the first ten minutes — you build a character, shoot a reel, and finish with something you can actually post. Plus a bonus track on reading your own social stats inside the platforms.
Every date is face-to-face at Paya Lebar Square (Office Lobby 1) — a small room, your own laptop, and Andrew moving between desks for three and a half hours.
Singapore 409051. Directions are sent to you after you book.
Right on top of the Paya Lebar interchange, with the mall underneath for coffee before and lunch after.
Three and a half hours is a long time to be at a table. Proper seating, quick Wi-Fi, coffee and tea on tap.
We're being straight with you: this course has real prerequisites, and they cost real money. You will be building alongside us — not watching a demo — so the accounts have to be live before you sit down.
The director. Your agent — Claude or ChatGPT Codex, your choice — drives the whole pipeline: writing prompts, running generations, assembling the cut. A paid plan is what makes the tool connections possible. Bring whichever you already use.
The camera — and where you get Seedance 2.0, the newest and most powerful video model available today. Every frame you generate comes from here, and the credits are consumed live in class. One month is enough to finish the course and keep producing afterwards.
The score. We only touch it lightly, at the end — the free tier will comfortably cover everything you do in the session.
Yes — that's roughly US$40–50 in subscriptions before you even pay us. We'd rather say so on the poster than surprise you on the day. In exchange, you don't watch someone else make a film. You make one.
Andrew doesn't teach this pipeline from a slide deck. He runs it — Jasmin is his, built and posted on exactly the workflow you'll learn, and every awkward failure in this course is one he has personally hit and solved.
An award-winning trainer who turns fast-moving AI tools into skills people use the same week. No prior video, design or coding experience is assumed. Every session is live, step by step, and you finish having actually made something.
"I've attended the AI class taught by DR Andrew. He's a helpful, jovial instructor who is willing to teach. His class is interesting."
"Dr.Andrew is always my best tutor in AI. Five stars is not enough"
"Informative. Clearcut"
"Learned great AI tools"
"I've learnt using AI — Claude, Google Flow, Suno, CapCut. It was amazing. Mr Andrew is very helpful."
"He's knowledgeable, eloquent, patient and approachable, and his classes are well-paced so no one gets left behind."
"There were so many valuable takeaways packed into the session. I started with no prior knowledge, and by the end of just 3 hours, I was able to produce my own video."
"Andrew is one of the best trainers I've ever had the privilege of learning from. He consistently goes above and beyond for every student."
"A jam-packed 3 hours online class. Another well done course by Dr Andrew."
"One of the best trainers around."
"I've attended the AI class taught by DR Andrew. He's a helpful, jovial instructor who is willing to teach. His class is interesting."
"Dr.Andrew is always my best tutor in AI. Five stars is not enough"
"Informative. Clearcut"
"Learned great AI tools"
"I've learnt using AI — Claude, Google Flow, Suno, CapCut. It was amazing. Mr Andrew is very helpful."
"He's knowledgeable, eloquent, patient and approachable, and his classes are well-paced so no one gets left behind."
"There were so many valuable takeaways packed into the session. I started with no prior knowledge, and by the end of just 3 hours, I was able to produce my own video."
"Andrew is one of the best trainers I've ever had the privilege of learning from. He consistently goes above and beyond for every student."
"A jam-packed 3 hours online class. Another well done course by Dr Andrew."
"One of the best trainers around."
None. The whole point of an automated pipeline is that you describe what you want and the tools do the labour. If you can write a clear sentence, you can direct a scene.
No — this is v2.0, and it's a different course. The original teaches you to make a video by hand with Google Flow, Suno and CapCut. This one drops the manual tools entirely and teaches you to build a character and run an automated production pipeline with your agent and Higgsfield. If you've done the first one, this is the sequel. If you haven't, you don't need it.
The masterclass is S$240 per seat — three and a half hours face-to-face with Dr. Andrew Chew, hands-on, with follow-up query support afterwards, and 10% back in points towards your next class. Book your date in the Reserve Seat portal and the joining details arrive by email. Note that your agent and Higgsfield subscriptions are separate from the course fee.
Claude Pro or ChatGPT Plus — whichever you already use, from US$20/mo — and Higgsfield (one month minimum; this is where your generation credits come from). Suno's free tier is enough. Both subscriptions can be cancelled after the month. We'd rather be blunt about this now than have you discover it on the day.
Yes. You build your character in Part One, generate a full sequence in Part Two, and in Part Three the agent assembles and scores the cut. You leave with a finished piece — and, more importantly, with the pipeline that made it.
Not at all. A consistent character is just as useful for a product advertisement, a brand campaign, a corporate explainer or a short film. Jasmin is the demonstration, not the assignment.
Always live and always face-to-face — three and a half hours in the room with Dr. Andrew Chew at Paya Lebar Square (Office Lobby 1), deliberately small so Andrew can reach every desk more than once, with follow-up query support afterwards. There is no online version of this class: the pipeline is involved enough that you want someone able to look at your actual screen.
Three and a half hours in the room. One character who never breaks. A pipeline that keeps working long after you walk out.
Reserve Seat. Book and Manage Courses.
