NGT Student kit · SFU

Generative AI workflow for video · Saint Francis University

Everything for today, in one place.

Sign-ups, the live poll, the submission folder, and every link from the slides. Keep this page open for the whole session. 今日所有連結都喺度,開住呢頁就得。

Date29 Oct 2026
FormatTalk + workshop
You makeA 3-shot film
Live poll · MentiJoin the voteWarm-up, questions, and the screening vote.menti.com/alkh92afqg3r → Hands-on 3 · SubmitUpload your filmMP4, named with your names.Open Google Drive folder →

01Do this now · 3 min

Open three accounts開三個帳號

You need all three before Hands-on 2. One account per pair is fine.

01Gemini (Nano Banana)Keyframes and images · free daily limit · any Google account
Open
02Doubao 豆包Video with Seedance 2.0 · about 5 free clips a day · sign in with Apple ID or a +852 phone. Tick the agreement box first.
Open
03CapCutEdit, music, titles · free
Open

Other video tools · if Doubao won't work

Google Vids10 free Veo 3.1 clips a month on a personal Google account
Open
Kling66 free credits a day · log in daily, they expire
Open
QwenWorkAlibaba's AI agent with Wan 3.0 · web and desktop
Open

Pro tools · worth knowing

HiggsfieldOne studio for many video models, with ready-made camera moves and effects
Open
GPT Image 2.5OpenAI's image model in ChatGPT · keyframes, posters and product shots
Open

Stuck? Raise a hand, or pair with a neighbour.

02One project · three stops

Your 3-shot film三次落手

Pick one brief: a drink or snack you invent, a sneaker or accessory you invent, a campus event or design show, or your own idea.

Step 1 · 7 min

Story, look, storyboard

  1. 0–2 minOne-line story: before → moment → after
  2. 2–4 minYour look: 3 colours + one style reference
  3. 4–7 min3-shot storyboard: shot size + action per shot
Step 2 · 12 min

Keyframes, then clips

  1. 0–5 minGemini: hero image first, then 3 keyframes. Same palette and style line. Leave room for your title.
  2. 5–9 minDoubao: each keyframe → video. Prompt the motion only. Start all three at once.
  3. 9–12 minOne more take of your hardest shot.
Step 3 · 15 min

Pick, fix, edit, submit

  1. 0–4 minWatch every take. Keep the best one per shot.
  2. 4–8 minFix the worst take. Change one thing. Re-roll once.
  3. 8–15 minCapCut: trim, music, end title. Export MP4 to the Drive folder.

03The formula to remember

5W1H for video prompts影片 prompt 公式

Same order every time. Who, what and where first; then camera, light and style.

WHOSubject

2–3 details you can see

WHATAction

One action, in beats

WHEREScene

Place and props

WHENLight

Time of day sets the light

WHYMood

The feeling, in two words

HOWCamera

One move · your palette · one sound

Example prompt

A matte pastel-green tea can with a silver ring-pull. Loose papers swirl up around it. A cluttered dorm desk. Early morning, soft window light from the left. Calm in the chaos. Wide, eye level, slow push-in, pastel palette, paper rustle.

Image → video in Doubao: describe the motion and camera only. The image already has the look.

04Prompt library · from the slides

Prompts you can copy簡報入面嘅 prompt,一撳複製

Every prompt shown in class. Swap in your own subject, keep the structure.

Video

Text-to-video · one shot

A matte pastel-green tea can stands on a cluttered dorm desk, loose papers swirling around it in the air. Wide shot, eye level, 35mm lens, shallow depth of field. Soft early-morning window light from the left. Clean, minimal product-film style, pastel palette. Calm but busy mood. Slow push-in toward the can. 5 seconds.

Image-to-video · motion only

Papers drift slowly around the can. Camera slowly pushes in. The can stays still and sharp.

With a keyframe, the image carries the look. The prompt carries the motion.
Timestamp · several shots in one 8 s clip

[00:00–00:02] Close-up, a cold can on a wet stone counter, drops sliding down.
[00:02–00:04] Macro shot, a finger pulls the tab. SFX: crack and fizz.
[00:04–00:06] Medium shot, she takes a sip and smiles, warm sunset light.
[00:06–00:08] Wide crane shot up to the skyline. A soft synth chord swells.

Works in Veo 3.1, Kling multi-shot and Seedance multi-shot. Say what you want, not what you don't: "an empty street", not "no people".
Concept film · one beat of seventeen

06.50–07.70: YOTTA takes one natural step forward. She raises one hand toward the viewer. Instead of a mechanical interface appearing suddenly, a thin translucent layer of light gently grows from her fingertips. Minimal. Transparent. White and cyan. 85mm close-up on her hand and face. Soft sunlight + subtle cyan rim light.

Time, action, what it is NOT, look, camera, light. Seventeen beats like this make the NGT concept film.
Concept film · the full prompt · 1,400 words, 17 beatsShow
The prompt behind the NGT concept film, word for word. <<<image_1>>> is the YOTTA reference image.

Create a 20-second premium cinematic brand film for NGT (Next Gen Tech) and its virtual AI companion character YOTTA <<<image_1>>>. CORE BRAND IDEA: “YOTTA, NGT — 引領未來,建造你所想。” English concept: “YOTTA. NGT. Lead the Future. Build What You Imagine.” The film should feel like a world-class premium future technology brand commercial, combining: Bright Futurism, Human-Centered Technology, Friendly AI, Optimistic Futurism, Clean Tech, Soft Sci-Fi, Smart City, Tech Utopia, Premium Futuristic design, Japanese-inspired futuristic minimalism. IMPORTANT: This is NOT cyberpunk. Do NOT use dark dystopian environments, aggressive neon colors, excessive holograms, robotic humans, mechanical armor, excessive circuitry, Matrix-style green code, or generic AI visuals. The future should feel: bright, clean, sophisticated, breathable, elegant, warm, optimistic, human, technologically advanced and emotionally inspiring. YOTTA: <<<image_1>>> A beautiful young adult East Asian woman, realistic human appearance, sophisticated and intelligent, energetic but composed, warm natural smile, confident eyes, youthful and professional. She should feel like a real human AI companion from a higher-dimensional future civilization, not a robot and not a conventional virtual idol. Appearance: long dark brown-black hair with subtle cyan-blue highlights, natural movement, refined facial features, intelligent bright eyes. Wardrobe: premium white futuristic techwear, translucent lightweight outer layer, silver micro-details, subtle cyan luminous accents, elegant minimalist silhouette, premium materials, no heavy armor. A small subtle Y-shaped luminous interface element is integrated into her clothing/accessories. Her technology feels invisible and elegant rather than mechanical. VISUAL LANGUAGE: white, silver, ice blue, cyan, soft navy accents. Natural daylight mixed with subtle futuristic illumination. Volumetric sunlight, atmospheric haze, glass reflections, soft bloom, realistic skin texture, realistic fabric, physically accurate materials. Premium commercial cinematography. Large-format cinema camera aesthetic. 35mm / 50mm / 85mm lens language. Shallow depth of field when focusing on YOTTA. Smooth controlled camera movement. No handheld shakiness. No excessive lens distortion. ================================================== MICRO TIMELINE — EXACT 20 SECONDS ================================================== 00.00–00.80 BLACK SCREEN. Almost complete silence. A single tiny cyan-white point of light appears in the center of darkness. The point pulses once, extremely subtle. Sound: one soft crystalline electronic pulse. Camera: static, perfectly centered. Mood: mysterious, elegant, premium. -------------------------------------------------- 00.80–01.70 The tiny point expands into an extremely thin luminous Y-shaped geometric light structure. The structure is not a logo yet. It feels like an advanced dimensional gateway. Soft white-blue particles begin floating around it. The surrounding darkness gradually becomes a bright atmospheric sky. Camera slowly pushes forward. Sound: deep soft sub-bass + delicate crystalline resonance. -------------------------------------------------- 01.70–03.00 The camera passes through the luminous structure. MATCH CUT into an enormous futuristic smart city. Reveal: a beautiful clean utopian city under a clear blue sky. Architecture: elegant glass towers, white architectural surfaces, vertical gardens, clean elevated transportation systems, autonomous vehicles, intelligent infrastructure, water reflections. The city feels like a believable future 20–30 years ahead. Not fantasy architecture. Not dystopian. Highly sophisticated and realistic. Sunlight floods through the buildings. Camera performs a graceful aerial forward movement between the towers. -------------------------------------------------- 03.00–04.20 The camera accelerates slightly through the city. Micro glimpses of future life: autonomous electric transportation moving silently, smart glass buildings adapting their transparency, people interacting naturally with invisible AI systems, clean energy infrastructure, green vertical gardens, subtle transparent interfaces embedded naturally into the environment. Technology is integrated into everyday life. No giant floating screens. No clutter. Everything feels effortless. Sound design becomes slightly more energetic. -------------------------------------------------- 04.20–05.20 The camera approaches a bright elevated terrace overlooking the city. A human silhouette stands in the sunlight. Wind gently moves her hair and translucent outer layer. We do not reveal her face yet. She stands naturally. Not posing like a model. She feels as if she has been waiting for us. Music begins to rise emotionally. -------------------------------------------------- 05.20–06.50 Slow cinematic orbit around YOTTA. Her face gradually enters the sunlight. Reveal her eyes. Natural skin texture. Natural facial micro-expression. Subtle smile. Her eyes briefly reflect the futuristic city. She looks directly toward the camera. This is the emotional connection point. VO, soft female voice: 「你想像過,未來會是什麼樣子嗎?」 -------------------------------------------------- 06.50–07.70 YOTTA takes one natural step forward. She raises one hand toward the viewer. Instead of a mechanical interface appearing suddenly, a thin translucent layer of light gently grows from her fingertips. Elegant holographic interface. Minimal. Transparent. White and cyan. Small particles respond to her movement. Camera: 85mm close-up on her hand and face. Lighting: soft sunlight + subtle cyan rim light. -------------------------------------------------- 07.70–09.00 The holographic interface expands. Inside it: a miniature digital model of a futuristic city. YOTTA moves her fingers naturally. The digital city responds instantly. Buildings rearrange. Transportation routes adapt. Energy flows illuminate. Green spaces expand. Everything happens smoothly and intuitively. VO: 「我是 Yotta,來自未來的智能體。」 -------------------------------------------------- 09.00–10.20 Extreme close-up of YOTTA's eyes. Inside her eye reflection: the digital city transforms into a real city. MATCH CUT. The camera emerges into the real city. The digital and physical worlds seamlessly become one. This transition should feel magical but scientifically elegant. -------------------------------------------------- 10.20–11.50 YOTTA walks through the futuristic city. People naturally interact with the technology around her. A child looks at an intelligent public installation. An elderly person receives assistance from a friendly AI interface. A designer manipulates a 3D architectural model. A worker interacts with an intelligent workspace. YOTTA is not controlling people. She is enabling them. This communicates: Human-Centered Technology. VO: 「NGT,不只是科技。」 -------------------------------------------------- 11.50–12.80 The environment begins transforming around YOTTA. Transparent data streams subtly connect: people → ideas → AI → technology → city. The visual language becomes more sophisticated. The connections form a large abstract luminous network. YOTTA walks through the center. She looks forward with confidence. VO: 「我們把科技,變成你的可能。」 -------------------------------------------------- 12.80–14.20 YOTTA stops. In front of her is an empty luminous architectural space. It represents an unrealized idea. She gently touches the air. A single point of light appears. From that point, an entire future environment rapidly but elegantly constructs itself: architecture, transportation, energy, AI systems, green spaces, human communities. The world is literally being BUILT from imagination. Camera pulls backward dramatically. -------------------------------------------------- 14.20–15.80 The entire futuristic city now surrounds YOTTA. She stands at the center. The city is bright, peaceful and alive. Sunlight breaks through clouds. A huge subtle luminous Y-shaped energy arc appears behind her. It feels like a symbol of possibility rather than a corporate logo. Music reaches emotional peak. VO: 「引領未來。」 Pause. -------------------------------------------------- 15.80–17.20 YOTTA turns toward the camera. She smiles slightly. She extends her hand. Not a dramatic superhero gesture. A simple human invitation. The luminous city behind her responds. Thousands of tiny lights activate progressively across the skyline. The future is coming alive. VO: 「建造你所想。」 -------------------------------------------------- 17.20–18.30 CUT TO PURE WHITE. A very subtle cyan luminous line draws across the screen. The line forms the elegant YOTTA wordmark. Minimal premium motion graphics. Text: YOTTA Underneath: NGT — NEXT GEN TECH Clean typography. Precise spacing. No excessive effects. -------------------------------------------------- 18.30–19.30 The YOTTA wordmark gently transitions into the NGT brand identity. Text: NGT NEXT GEN TECH Below: 「引領未來,建造你所想」 English: LEAD THE FUTURE. BUILD WHAT YOU IMAGINE. Background remains extremely clean white with subtle atmospheric blue gradients. -------------------------------------------------- 19.30–20.00 Final hero frame. YOTTA appears subtly in the background, standing inside the bright futuristic city. Foreground: NGT NEXT GEN TECH YOTTA 「引領未來,建造你所想」 Final sound: one clean crystalline tone. Fade to white. ================================================== CAMERA & CINEMATOGRAPHY ================================================== Premium global technology commercial. Large-format cinema look. Highly realistic human skin. Natural facial expressions. Elegant camera choreography. Slow controlled dolly movements. Smooth aerial cinematography. Subtle orbit shots. Macro details. 85mm portrait compression for YOTTA. 35mm environmental storytelling. Occasional 50mm cinematic perspective. Use: soft natural sunlight, volumetric light, subtle atmospheric perspective, realistic reflections, soft optical bloom, controlled highlights, clean shadows, cinematic depth of field. No excessive slow motion. No shaky camera. No cheap sci-fi visual effects. ================================================== YOTTA PERFORMANCE ================================================== YOTTA should behave like an intelligent human companion. She is: energetic, curious, warm, precise, confident, optimistic, smart, friendly. Her movements are subtle and believable. Avoid: idol-like exaggerated gestures, seductive posing, robotic movements, superhero poses, overacting. She should communicate: “I understand what you imagine, and I can help you build it.” ================================================== SOUND DESIGN ================================================== Start almost silent. 0–3 seconds: minimal crystalline electronic pulse. 3–7 seconds: soft futuristic atmospheric soundscape. 7–12 seconds: subtle intelligent digital textures. 12–17 seconds: emotional cinematic music gradually rises. 17–20 seconds: music resolves into a clean premium sonic logo. Music reference: modern cinematic technology brand, minimal orchestral texture, soft piano-like tonal elements, subtle electronic sound design, emotionally uplifting, not aggressive EDM. ================================================== OVERALL EMOTIONAL ARC ================================================== MYSTERY → WONDER → HUMAN CONNECTION → INTELLIGENCE → POSSIBILITY → CREATION → OPTIMISM → BRAND The audience should finish the film feeling: “The future is not something far away.” “It is something we can imagine.” “And NGT can help us build it.” FINAL BRAND MESSAGE: YOTTA. NGT — NEXT GEN TECH. 引領未來,建造你所想。 LEAD THE FUTURE. BUILD WHAT YOU IMAGINE.

Watch the film it made ↗ · Brand film for NGT and YOTTA. Learn from the structure; write your own brand and character.

Look and design

Prompt like an art director

Tea can on the right third; top third left empty for the title. Palette: sage green, off-white, warm grey. Soft key light from the left, low contrast. Matte can, recycled-paper texture. Clean Swiss-style layout. Slow push-in.

Before: "A tea can on a desk, papers flying, cinematic, beautiful, 4K." Cut the empty words; name 3–5 colours.
Style bible · paste into every prompt

[CHARACTER] YOTTA: blue twin-tail hair, white-silver suit, cyan glow lines · ref: character-sheet.png
[BRAND] night navy #060A13, icy cyan #4FD8EB, orange #FF7A45 accents only
[STYLE] blue hour, pearl-white towers, soft bloom, 35mm · ref: style-board.png
[SHOT] ← the only line you change

Words keep the style. Reference images keep the face, the product and the logo.
Fix one thing · real NGT project

Make ONE narrowly scoped change: constrain every decorative symbol to the official BLUE palette. Keep everything else exactly the same.

Name the one change. Then list what must not move.
Chatbot as prompt partner

You are a cinematographer. Here is my storyboard row: [subject, action, environment, framing, camera, 5s]. Write 3 different image-to-video prompts. Each one: one camera move, one main action, what must stay still. Under 60 words each.

Music and coded video

Suno · custom mode

Style of music: cinematic electronic, dark synth pads, punchy drums, 110 BPM, no vocals
Lyrics: [Intro] [Build] [Drop] [Outro]

Instrumental on, make 2 versions. At 120 BPM a beat lands every 0.5 s: cut your shots on the beat.
Claude · your first coded video

Make a 15-second 9:16 motion graphic as ONE HTML file.
Draw everything on a <canvas> in renderFrame(t), t in seconds.
The same t must always draw the same picture.
Look: dark background, one accent colour, bold sans type.
0–4 s: hook "…". 4–11 s: three facts, one per beat. 11–15 s: logo.
Add play, pause and a scrubber so I can preview it.

Preview it in Claude straight away. For an MP4, ask Claude Code to render 30 fps with ffmpeg, or screen-record the preview.
Claude Code · find the beats

Analyse music/track.mp3. Find the tempo, every beat, each bar and where the energy changes. Save it as beats.js.

Music first. The beat map tells Claude when each scene starts.

05Hands-on 3 · troubleshooting

When the take goes wrong出錯點救

ProblemTry this
Subject morphsAdd a reference image; add "stays unchanged"; shorten to 5 s
Unnatural motionDescribe one simple action; use slow, gentle verbs
Camera too wildName one move; add "slow" or "steady"; try static
Nothing movesName what moves and how: hair, steam, fabric, light
Shots don't matchSame hero reference and style line in every shot
Text or logo garbledRemove it from the image; add it in the edit

Edit tips

Trim hardCut the first and last half-second.
Cut on motionHard cuts beat fancy transitions.
Pace to musicLay the track first. Cut on the beat.
Type itProduct name and one line on the last shot.
Add soundOne or two effects make it feel real.

06Soundtrack

Type a mood, get a songAI 配樂 · Suno

SunoFree: 50 credits a day (about 10 songs on v6-mini). Switch on Instrumental for background music. Free plan is personal use only, and downloads are limited.
Open

Can't download from Suno? Use the built-in music in CapCut for today.

07Student budget · start at $0

Make your film for free零成本做片

Stack daily creditsKling gives 66 a day; Doubao gives free Seedance clips daily. Log in every day.
Free Veo in Google Vids10 Veo 3.1 videos a month on a personal Google account.
Test cheap, render onceDraft at 720p or in a Fast/Lite model. Only the final take goes to 1080p.

Free tiers add watermarks and ban commercial use. For client work, pay.

08Research · official, 2026

Prompt guides from the labs官方指南

Google · Veo 3.1Ultimate prompting guide
Read
Kuaishou · KlingText-to-video prompt guide
Read
ByteDance · Seedance 2.5Six-part formula, with examples
Read
OpenAI · Sora 2Prompting guide (cookbook)
Read
StudioBinderHow to make a storyboard
Read

09NGT work · seen in the talk

Made by NGT, with AI講座入面嘅 NGT 作品

The YOTTA films and coded videos from the slides. Watch them again and take apart the shots.

YOTTA concept filmAI concept film · Instagram
Watch
YOTTA game promoGame trailer · Instagram
Watch
AI-coded video · 1Motion written as code by AI · Instagram
Watch
AI-coded video · 2Motion written as code by AI · Instagram
Watch
Video inside a websiteAI video built into a live site · yotta.ngt.hk
Visit

10Keep going

Where to go after today課後延伸

Style Atlas200 image styles on one theme, every prompt included
Open
Motion EngineeringHow pros turn one prompt into a 10-step system
Open
Agent skillsReady-made workflows for UGC ads and product motion
Open
AI short-drama canvasScript → characters → shots → video, run by an agent
Open
AI-coded motionA 30-second launch film from one prompt
Open
Your libraryEvery week, take apart 3 shots from one film you love.
Habit

Some links are commercial platforms or promo posts. They are references, not recommendations to buy.