- The Shift
- Posts
- Can you pass the AI video test?
Can you pass the AI video test?
Plus, 🎨 How to turn Figma designs into interactive prototypes + code with Figma Connect, AI music gets a “real” release, and more!
In partnership with
Welcome back to The Shift. Let’s get straight to what matters in AI today…
Today we have:
🎥 Most People Can’t Spot AI Video Anymore
🎨 How to Turn Figma Designs Into Interactive Prototypes + Code with Figma Connect
🎵 AI Music Gets a “Real” Release
🔨Tools and Shifts you cannot miss
🎥 Most People Can’t Spot AI Video Anymore
Runway just published a reality check on how believable AI video has become. It tested whether everyday viewers can tell a real clip from one generated by its Gen-4.5 model. The results suggest we’ve crossed into a new phase where “just use your eyes” is no longer enough.
The Shift:
1. Runway Tested Real vs AI Clips - Runway recruited 1,043 people and showed each person 20 five-second videos. Half were real, and half were generated by Gen-4.5, shown in random order. The AI clips were not edited or regenerated to improve quality.
2. Detection Was Barely Above Chance - Over 90% of participants could not reliably tell the difference. Only 9.5% scored high enough to be statistically accurate, getting at least 15 out of 20 correct. Overall accuracy was 57.1%, close to guessing.
3. Some Content Is Harder Than Others - People did slightly better on human-focused clips like faces, hands, and actions, landing around 58–65% accuracy. But animals and architecture dropped below chance at 45–47%.
Runway says we’ve hit a tipping point where “spot the fake” is no longer a real safety net, especially when synthetic clips can look more believable than reality. Gen-4.5 now tops Artificial Analysis’ text-to-video rankings, with new image-to-video rolling out this week.
TOGETHER WITH BEEHIIV
Help us make better ads
Did you recently see an ad for beehiiv in a newsletter? We’re running a short brand lift survey to understand what’s actually breaking through (and what’s not).
It takes about 20 seconds, the questions are super easy, and your feedback directly helps us improve how we show up in the newsletters you read and love.
If you’ve got a few moments, we’d really appreciate your insight.
🎨 How to Turn Figma Designs Into Interactive Prototypes + Code with Figma Connect
Figma Connect lets you copy a Figma design and paste it into MagicPath to instantly turn it into a living, interactive prototype. No plugins, no MCP setup, no pixel loss.
How to use it
Import from Figma: Open MagicPath and click Import from Figma, or start from the homepage.
Connect your Figma account: One-time connect, then you’re ready to move fast.
Copy your design in Figma: Press ⌘L on Mac or Ctrl + L on Windows.
Paste into MagicPath: Press ⌘V and your layout, typography, colors, and assets stay perfectly preserved.
✨ Now you can edit with AI using your design system, share interactive links, and export production-ready code. You can try it here.
🎵 AI Music Gets a “Real” Release
ElevenLabs just dropped an AI-made album featuring Liza Minnelli and Art Garfunkel, now live on Spotify and ElevenLabs’ site. It’s a clear play to make AI music feel artist-first and commercially safe. Meanwhile, Adobe is building private, IP-safe Firefly models for studios.
The Shift:
ElevenLabs Turns AI Music Into a Showcase - ElevenLabs released “The Eleven Album” to promote Eleven Music and its Iconic Voices Marketplace, framing it as proof that artists can use AI to expand creativity while still keeping full authorship and commercial rights.
ElevenLabs says each artist created a fully original track, retains full ownership, and earns 100% of streaming revenue.
Adobe Goes Studio-First - Adobe is developing Firefly Foundry “private, IP-safe” omni-models for entertainment production, designed to generate assets like audio-aware video and 3D or vector graphics that slot neatly into workflows like Premiere.
Instead of scraping the open internet, Adobe says each model is trained only on IP the studio actually owns, so teams can move faster from pre-vis to final edits while protecting creative intent and avoiding ownership fights.
AI music is trying to win trust by giving creators ownership and payouts, while Hollywood genAI is heading toward private models that keep everything legally clean. Different strategies, same message: the future is AI that can ship commercially without drama.
TOGETHER WITH WISPR FLOW
Better input, better output
Voice-first prompts capture details you forget to type. Wispr Flow turns speech into clean prompts you can paste into your AI tools for faster, more useful results. Try Wispr Flow for AI.
🔨AI Tools for the Shift
🎙️ Todoist Ramble – Capture tasks at the speed of thought, just talk naturally and Ramble turns your voice into clean, organized to-dos instantly.
🗓️ Blockit AI – Your 24/7 AI time assistant that actually works, helping you schedule faster and remove the back-and-forth that kills momentum.
🎥 VidFlux – Turn any static photo into a cinematic AI video in seconds, adding realistic motion, depth, and dynamic storytelling to your best moments.
🏟️ SportBot AI – Compare AI win probabilities vs market odds so you can spot mispriced bets and make sharper decisions without guessing.
📩 Enquiry Genie – Generate AI-powered email replies in seconds and get real-time ranking alerts when your pricing starts hurting visibility.
🚀Quick Shifts
💻 Claude Code is gaining major traction inside Microsoft, with teams across CoreAI, Windows, and Microsoft 365 being encouraged to use it alongside GitHub Copilot, even pushing non-coders to prototype faster.
🏛️ Lawmakers introduced the bipartisan TRAIN Act to let creators check whether AI models were trained on their copyrighted work, backed by groups like the RIAA and SAG-AFTRA.
🔍 Google’s Search AI Mode is adding “Personal Intelligence,” letting opted-in users connect Gmail and Google Photos so results can reflect bookings, past purchases, and personal habits, while Google says it won’t train directly on your inbox or photo library.
🔍 Google DeepMind CEO Demis Hassabis says he’s surprised OpenAI is moving so quickly to test ads in ChatGPT, warning that ads could weaken trust in AI assistants compared to Search, and says Google has no current plans to add ads to Gemini.
🎙️ Google DeepMind is hiring Hume AI’s CEO and key engineers through a licensing deal to boost Gemini’s voice features, signaling voice and emotion-aware AI as the next major battleground for assistants and wearables.
🧩 Prompt of the Day
How to Build an Influencer Partnership Plan Using One Prompt
Turn random creator outreach into a clear blueprint for collaborations that drive trust, content, and conversions.
Paste the prompt: Drop this into ChatGPT, then fill in your ideal influencer niche, audience size, and what content you want them to produce.
Prompt to paste
Create an influencer partnership blueprint for [Brand]. Include influencer criteria (niche, audience size, engagement quality), deliverables (content types and quantity), and compensation structure (flat fee, gifting, commission, or hybrid). Keep it practical, clear, and optimized for authentic endorsements.
That’s all for today’s edition see you tomorrow as we track down and get you all that matters in the daily AI Shift!
If you loved this edition let us know how much:
How good and useful was today's edition |
Forward it to your pal to give them a daily dose of the shift so they can 👇
Reply