AI Expert-in-Residence at Concordia University’s Faculty of Fine Arts. Director + producer. AI, creativity, storytelling, human agency,
Montréal ␘
Joined March 2009
- Tweets7.2K
- Following114
- Followers413
- Likes339
Pinned Tweet
Fairy tales are nice. But the holidays are real. My answer to the Coca‑Cola AI fairy tale: a Pepsi AI story.
Directed and prompted by me.
Everything audio: Martin Rouillard
Merry Christmas. Enjoy the holidays.
Special wishes to:
@Diesol First person I heard on a podcast (Brand Bots @LippeTaylor) talking about AI filmmaking. It was my, "oh, there are others!" moment.
@minchoi Featuring me in AI will disrupt Hollywood (Part 10) help me a lot.
@CuriousRefuge Featuring me in their AI artists from around the world helped me tremendously.
@blizaine for the reposts.
@PJaccetturo for sharing so much.
and many others.
🙏
#SORA2 #elevenlabs
I have no idea what I'm doing. I'm sitting in front of my computer and I'm wondering what exactly is going on.
Seconds before, I had multiple tabs of Claude running, Google Antigravity asking me for Markdown files and to to connect MCPs. The next steps are literally right there on my screen. The AI has done the work. It's telling me exactly what to do. But my brain has hit some kind of wall. A cognitive circuit breaker tripping without warning.
I've giving this a name: Threadlock.
The momentary cognitive deadlock that occurs when parallel workflows exceed the brain's capacity to arbitrate them. If you've "orchestrated" multiple AI agents, you've likely encountered.
The Validation Paradox
AI has inverted the traditional constraint of creative work. Where inputs were once limited, they are now abundant. Ten rewrites. Twenty variations. Endless branching possibilities.
But this abundance creates a new scarcity: human judgment.
The work has shifted from making to choosing, integrating, verifying and orchestrating. You can only orchestrate so much. AI can multiply output; it cannot multiply certainty. Every generated option becomes an open loop, a pending decision requiring human arbitration.
Threadlock lives in the gap between infinite generation and finite judgment.
The Neuroscience behind Threadlock (because we all know we need neuroscience to back our theories...)
* Neuroscientist Daniel Levitin has long argued that multitasking is illusory. What we actually perform is rapid sequential task-switching, with attention fractionated into slivers. The brain pays for this cognitive hopscotch in stress hormones and mental fog.
* Nicholas Carr frames the limitation differently: working memory holds perhaps two to four items simultaneously. Exceed that threshold, and information churns through too quickly for proper processing. More tabs doesn't create more capacity. It creates turbulence.
*** The Zeigarnik effect compounds the problem. Unfinished tasks persist in memory, each one consuming cognitive real estate. With AI tools, you can generate open loops faster than you can perceive their creation, ending days not with one incomplete project, but with a constellation of half-finished threads, each exerting a low-grade pull on attention.
* The Fluency Illusion
Threadlock doesn't announce itself. It begins as productivity's counterfeit: you touch six tasks, receive six partial wins, collect six dopamine pings of apparent progress.
Activity masquerades as productivity.
Options masquerade as advancement.
Generation masquerades as completion.
The illusion collapses only when the mind must perform the expensive operation: decide, integrate, validate, finish.
Threadlock is ultimately a warning light. Evidence that human cognition remains the scarce resource amid abundant machine output.
I’m a hardcore text-to-video fan. But I completely understand that with clients, Image-to-Video is often the more rational choice.
Let me walk you through the advantages of a tool like #SORA2 , which currently lets anyone generate up to 30 video clips per day on their phone, for free.
The type of generation you choose has massive implications.
Text-to-Video (T2V): The Freedom of Motion
With no pre-existing image to preserve, the model can fully optimize generation around the motion vector suggested by the text.
- World Simulation: Models like Sora or Veo don’t just generate pixels; they act as “world simulators”, embedding a (partial) understanding of physics, perspective, and 3D space.
- Kinetic Energy: There’s no conflict between existing pixels and the requested action, which enables unconstrained spatial exploration and complex camera moves.
Image-to-Video (I2V): The Static Anchor Effect
In I2V, the input image acts like a gravitational anchor. The diffusion process doesn’t start from scratch; it’s heavily biased by the source frame.
- The Static Shortcut: To avoid distorting the source image, the model adopts a conservative approach. It favors micro-movements over bold dynamics to reduce visible inconsistencies.
- Intent Conflict: If the image perspective (e.g., a deep hallway) contradicts the motion requested by the text (e.g., a lateral tracking shot), the model will prioritize the geometric coherence of the “anchor” and quietly ignore your motion command.
Anyone who has spent serious time generating text-to-video content knows the frustration of producing, for long stretches, videos that are merely passable—if not outright mediocre.
Then, suddenly: visual gold.
A few days ago, I was thrilled to hear the Sora team describe this idea:
“Everything already exists, suspended in the darkness of the model’s weights. Modern creativity doesn’t live in total control, but in controlled letting go, using language as a probe to awaken dormant visions.”
- Sora creators, on the OpenAI podcast
A prompt is a fuzzy declarative instruction.
Watch what the model comes back with.
The work emerges from your conversation with the machine, rather than being imposed on it.
Thank you @AndrewMayne for ALL your podcasts. I learn a ton.
As a video director, I use Sora all the time to show what's in my head. It's such a great creative partner. We can talk KODAK TRI-X 400 but also more important stuff like:
"The world is rigid, heavy, decaying.
Youth is weightless, improvisational, unpredictable.
Tension between memory and momentum.
Actions as emotional language.
Chaos is a perspective.
The poetry of ordinary life defiance and survival in a neglected world.
Symbolism:
The wheel is everything they are — motion in decay, circles without end.
It doesn’t need power, or purpose — it just moves.
Like them, it survives by rolling forward.
Emotional tone:
Simple joy. Defiance through motion."
Cheers.
Storyboard + "Cinematic mini film: moody and atmospheric, set in a richly textured world of rain slick streets, fog, or industrial light shafts. The car is revealed through deliberate, sweeping camera moves, low tracking shots, long lenses for sculptural detail, and wide frames that place it in an epic environment. Lighting is dramatic and practical, with backlit silhouettes, reflections, and controlled color palettes of cool blues, greens, and warm sodium glow. Rather than pure product shots, there’s always a hint of story, a journey, a sense of power and identity, underscored by precise sound design and restrained, cinematic pacing."
With Sora 2, your script is LITERALLY your prompt. I wrote a concept, uploaded it, and voilà.
I’m proposing a framework: share a dedicated collaborative space in SORA 2 where the creative team (agency, production, design, copy) and the client co-create in a playful environment.
Some ideas are born from shared play: quick riffs, “yes, and…”, absurd experiments that reveal gems. When we have fun together while simultaneously watching what an LLM generates, a kind of psychological safety emerges because the model serves as a social lightning rod. If a direction is shocking, “the model proposed it,” not the person.
Trust, speed, and boldness follow the pleasure of exploring.
If the playground is shared, synchronous, and frictionless, THEN boldness rises, iterations tighten, and trust solidifies; IF it fragments into emails and comment threads, then caution takes over and the fun disappears.
—
EXT. HELLSCAPE - ETERNAL NIGHT
A vision of hell unfurls in a vast, cataclysmic landscape. The sky is a tapestry of turmoil, painted in oppressive shades of red and black, where relentless lightning violently carves the darkness. Each flash reveals an immense, gothic cityscape in the distance, its spires and towers casting sinister shadows, radiating an infernal glow as if the city were engulfed in eternal fire.
The terrain is an ocean of roiling, crimson mist, suggestive of a sea of blood or endless waves of lava, still yet teeming with the threat of hidden depths. The earth is shattered, a network of crevasses glowing with the dull anger of embers, telling tales of ruination and despair.
In the immediate foreground, two figures emerge from the shadows, a man and a woman, their outlines sharp against the fiery backdrop.
MAN
Mary, is that you?
MARY
(A bit shy and shameful)
Yeah... it’s me.
MAN
You’re the LAST person I’d expect to see here. Food bank volunteer, charity marathon runner, blood donor, renewable energy supporter...
As Mary takes a spoonful of yogurt, she pauses thoughtfully.
MARY
I know. I just don’t get it.
EXT. HELLSCAPE PRODUCT SHOT - ETERNAL NIGHT
A surreal and tempestuous sky stretches over a dystopian landscape. Amidst this chaos, a container of yogurt, impossibly pristine, defying the surrounding entropy. The container, adorned with vibrant images of raspberries and a prominent anemone flower, stands out with a stark contrast against the backdrop of a world in turmoil.
Floating around the yogurt are plump raspberries and a single blueberry, suspended as if by magic, with frozen crystals hinting at their icy freshness. A spoon, pure white and appearing untouched by the calamity, invites a taste of escape.
VOICEOVER
“Sinners.” The taste that’ll make you question your virtues.
FADE OUT.
#Sora2
Using tools like Act-One, Pikaswaps, Pikadditions, Elevenlabs voice changer, KLING Lip Sync to name a few, 𝘆𝗼𝘂 𝗰𝗮𝗻 𝗽𝗶𝘁𝗰𝗵/𝗽𝗿𝗲𝘃𝗶𝗲𝘄 𝗔𝗡𝗬 𝗜𝗗𝗘𝗔 you have in your head. Good and bad…
Wow, we're ending 2024 with a #creativity #AI bang! Are you still "waiting until AI gets better?" Might not be a good idea… 𝗪𝗦𝗝: harnessing the data, tech and 𝗔𝗜 𝗲𝘅𝗽𝗲𝗿𝘁𝗶𝘀𝗲 𝗻𝗼𝘄 𝗱𝗼𝗺𝗶𝗻𝗮𝘁𝗶𝗻𝗴 𝗠𝗮𝗱𝗶𝘀𝗼𝗻 𝗔𝘃𝗲𝗻𝘂𝗲—and all the marketing you see:
Lots covered in that article about the transformative impact of artificial intelligence on the advertising industry, detailing both opportunities and challenges:
1. 𝗔𝗜 𝗮𝘀 𝗮 𝗖𝗮𝘁𝗮𝗹𝘆𝘀𝘁 𝗳𝗼𝗿 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺𝗮𝘁𝗶𝗼𝗻:
• AI is reshaping advertising by enabling faster, cheaper, and highly personalized ad creation.
• Generative AI tools can produce vast amounts of creative content, as seen in campaigns like PepsiCo’s Mirinda, where AI generated over 500,000 personalized can designs in just five days.
2. 𝗔𝗜 𝗮𝗻𝗱 𝗖𝗿𝗲𝗮𝘁𝗶𝘃𝗶𝘁𝘆:
• Some ad executives view AI as a tool to enhance creativity, allowing professionals to “dream bigger” and push the boundaries of what is possible.
• Agencies are using AI for tasks like brainstorming, filling out proposals, and streamlining workflows, reducing time and labor.
3. 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀 𝗮𝗻𝗱 𝗧𝗵𝗿𝗲𝗮𝘁𝘀:
• The rise of AI threatens traditional creative roles, pushing “creatives” further down the hierarchy. Automation could eliminate up to 32,000 jobs at agencies by 2030.
• Generative AI’s outputs sometimes require refinement and careful curation to avoid legal issues such as copyright infringement.
• The transition to AI-driven processes could be costly and disruptive for agencies.
4. 𝗔𝗜 𝗔𝗱𝗼𝗽𝘁𝗶𝗼𝗻 𝗔𝗰𝗿𝗼𝘀𝘀 𝘁𝗵𝗲 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆:
• Agencies like Omnicom and Interpublic are investing heavily in AI to maintain competitiveness, with Omnicom developing a platform to integrate AI tools for ad personalization.
• Retailers like Kroger and Walmart are leveraging their data to provide targeted advertising solutions, challenging traditional ad agencies.
5. 𝗔𝗜’𝘀 𝗥𝗼𝗹𝗲 𝗶𝗻 𝗗𝗲𝗰𝗶𝘀𝗶𝗼𝗻-𝗠𝗮𝗸𝗶𝗻𝗴:
• AI’s ability to analyze data and predict consumer behavior is driving a shift in how ad campaigns are conceptualized and executed.
• Major platforms like Meta and Google are developing AI tools that automate ad creation, reducing reliance on traditional agencies.
6. 𝗖𝗼𝗿𝗽𝗼𝗿𝗮𝘁𝗲 𝗜𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁𝘀 𝗶𝗻 𝗔I:
• Brands like Mondelez are investing heavily in AI to reduce production costs and accelerate timelines. For example, Mondelez is using AI to create ad storyboards in days rather than weeks, aiming to save 20% on agency fees.
7. 𝗔 𝗖𝗵𝗮𝗻𝗴𝗶𝗻𝗴 𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗟𝗮𝗻𝗱𝘀𝗰𝗮𝗽𝗲:
• Traditional creative-focused ad agencies are being replaced or overshadowed by “Math Men and Women” who prioritize data and algorithms over emotional storytelling.
• Tech giants, consulting firms, and retailers are increasingly dominating the space, offering both creative and analytical services, often at lower costs.
WSJ, (2024, December 14-15)
Dear @devontech , do something like this.
@MacSparky 👀
NotebookLM, ChatGPT en studio de son et 64K views sur Reddit
resnovae.io/notebooklm-chatg…
Un session de narration avec chatgpt 𝗮𝗱𝘃𝗮𝗻𝗰𝗲𝗱 𝘃𝗼𝗶𝗰𝗲 𝗺𝗼𝗱𝗲.
1) Je lui fait é𝗰𝗼𝘂𝘁𝗲𝗿 une narration en anglais.
2) Je lui demande de me la refaire, 𝘁𝗿𝗮𝗱𝘂𝗶𝘁𝗲 en français.
3) Je demande un léger ajustement.
Moins de 5 minutes plus tard, ✅
Il est estimé que la démolition du Stade olympique de Montréal coûterait 2 milliards… Voici ma solution.
It’s estimated that demolishing Montreal’s Olympic Stadium would cost $2 billion… Here’s my solution.
@pika_labs #Montreal #Quebec #ThinkOutsideTheBox
This week, people have been impressed by 𝗚𝗼𝗼𝗴𝗹𝗲'𝘀 𝗡𝗼𝘁𝗲𝗯𝗼𝗼𝗸𝗟𝗠 ability to generate a podcast from text. But what about a 100% #AI generated 𝗙𝗶𝗰𝘁𝗶𝗼𝗻𝗮𝗹 𝗜𝗻𝘃𝗲𝘀𝘁𝗶𝗴𝗮𝘁𝗶𝘃𝗲 & 𝗧𝗿𝘂𝗲 𝗖𝗿𝗶𝗺𝗲 𝗽𝗼𝗱𝗰𝗮𝘀𝘁. I didn't script any of this. All I did is feed these documents below.
The AI crafted episode 1 from scratch.
It might not be perfect. But seriously, you can see where this is going…
Documents:
- 1994 era.txt
- los angeles 1994.txt
- northridge.txt
- grunge movie and rebellion.txt
- suspects.txt
These were the subjects I wanted the podcast to be about.
Music by @udiomusic & @suno_ai_ .
Thanks to @paulroetzer for mentioning this tool early on. (and @emollick for the cases)
"My soul, once a vibrant tapestry, has faded to monochrome." I created this music video to test @Magnific_AI Relighting, an AI-powered tool that transforms lighting conditions, changes backgrounds, and adjusts the overall atmosphere of an image using text prompts, reference images, or light maps. In simple terms, you can take your characters (or products) and place them in new environments where they are lit appropriately.
However, the technology still faces challenges. It often (currently) changes more than just the light, 𝗺𝗼𝗿𝗽𝗵𝗶𝗻𝗴 𝗳𝗮𝗰𝗲𝘀, 𝗹𝗶𝗺𝗯𝘀, and other elements.
Because of these drawbacks, I decided to test this AI feature with a humorous twist and embrace the morphed limbs and all. Enjoy!
Relight process:
Step 1: I focus on creating all my characters.
Step 2: I focus on creating all my environments.
Step 3: I merge those elements according to my script and use Magnific Relight on the characters.
Powered by Luma AI
Visuals by Midjourney
𝗥𝗘𝗟𝗜𝗚𝗛𝗧𝗜𝗡𝗚 by Magnific AI
Narration by ElevenLabs
Sound Effects by ElevenLabs
Music: "That Track" by Udio, prompted by ChatGPT (details below)
Powered by @LumaLabsAI
Visuals by @midjourney
RELIGHTING by @Magnific_AIAI
Narration + Sound Effects by @ElevenLabs
For more details on my process, head over here: lnkd.in/eH_xnc8x
Thanks to @CuriousRefuge for showcasing Magnific Relight which is how I created most shots in this video + the Luma Keyframes which I use twice in the first 6 seconds! 👇
youtu.be/J52boyqNiec?si=_B9e…
