Software was Monday. Today AI makes voices, films, and music.
You meet the media stack and make one real asset for your product before standup.
ISB AI FoundersMULTIMEDIA AI
The media stack
Four lanes. Pick the one your product needs first.
Voice
ElevenLabs
Design a voice. It reads anything you give it.
For you: your landing headline, read like a trailer.
Image
Ideogram / Midjourney
Logos, brand shots, app-store art.
For you: a real logo in five seconds.
Video
Kling / Runway / Pika
5-15 second clips from a text prompt.
For you: one shot of your product's big moment.
Music
Suno
A soundtrack from one line of instructions.
For you: the track under tomorrow's launch video.
Network reality: Kling (可灵) and Hailuo/MiniMax are China-native, they work on any laptop in this room right now. ElevenLabs, Midjourney/Ideogram, Runway/Pika and Suno generally need the VPN laptop, or you watch the presenter demo.
Lane 1 · Voice
Design a voice. Have it read your own headline.
Live demo · no rehearsal
I pick a voice in ElevenLabs, paste in a real headline from this morning's landing pages, and play it out loud, right now.
You do
Watch, unless your team has the VPN laptop. Steal the prompt pattern below, you will use it in ten minutes.
"A warm, energetic teenage narrator, slight excitement, reads: [your headline]"
Lane 2 · Image
Your logo. Your brand shot. Same adjective, every time.
Live demo · no rehearsal
I run a logo prompt, then a brand-shot prompt, for one real team, using their design-system adjective both times.
You do
If you're on a VPN laptop, run your own version now. Everyone else: steal these two patterns for the exercise.
"[product name] logo, [design-system adjective], flat vector, single color background"
"cinematic photo of [your person] in [the problem moment], [design-system adjective]"
Your design-system adjective from this morning's style card is the style key. Type the same word into every prompt and everything matches.
Lane 3 · Video
KLING · no VPN needed
One shot = subject + motion + camera + mood.
Live demo · no rehearsal
I run this exact LunchRush prompt in Kling on the school wifi, right now, and we watch the clip render.
You do
Open Kling now, everyone can. In the exercise you'll write this exact structure for your own product's moment.
"A student's phone lifts a pickup code to a cafeteria counter scanner, quick handheld push-in, lunchtime crowd blurred behind, bright and fast."
Kling tip: shorter prompts win. One action per shot, 5-10 second clips, do not ask for a whole story in one go.
Lane 4 · Music, and the assembly truth
Tonight is ingredients. Tomorrow is the meal.
Suno needs one line
"upbeat, confident, 30 seconds, no vocals"
Shots
→
Voice
→
Music
→
CapCut, tomorrow
Every asset you make tonight is a piece of tomorrow's 30-60 second launch video. Nobody assembles anything tonight; that is the whole point of Day 4's production block.
Copy this prompt · run it in Claude or ChatGPT
The shot-list prompt.
"Write prompts for a 15-second launch clip for my product: [thesis]. Give me: a shot list of 3 shots with camera movement, a one-sentence video-generation prompt per shot in a cinematic style, a 20-word voiceover script reading naturally, and a one-line music brief. Match this mood: [your design-system adjective]."
Exercise · TEAM
10 min · lands by 16:12
One asset. Now.
1Pick one lane: voiced headline, logo, one Kling clip, or a shot list from the prompt above.
2Make it. Save the asset AND its exact prompt to your team folder.
3One kanban card: name, asset, done by 16:12. Early finish? Make a second lane's asset.
10:00
Asset + prompt, team folder
16:12 · standup in three
LEAVE WITH · one real asset, plus its exact prompt, saved in your team folder.
DECIDED · which lane your launch video leans on.
NEXT · standup, right now: asset on screen next to your fastest fix. Tomorrow 13:15, S10 The launch video: script, voice, shots and music become your launch video.