Hey, it's the Carat team. 🤗 Today, we're looking at an AI World Cup campaign with nearly 100 million views, a 4K recreation of life 900 years ago, Claude's values across languages, and a workflow that transfers Blender camera motion to Seedance 2.
Kling AI World Cup campaign nears 100M views
ⓒ @KlingAIOfficial
A Portugal World Cup campaign made with Kling AI has pulled nearly 100 million views on Instagram. It combines live action, traditional VFX, 3D, and AI in a hybrid production.
The film follows Portugal players and fans, then joins the waters off Lisbon and New York into a single coastline near the end. According to Kling AI, around 80% of the film was created with AI, with Kling serving as the primary generation tool.
📌 Today's 3 stories
1️⃣ AI recreated life 900 years ago in 4K
ⓒ @umesh_ai
AI creator Umesh recreated a day in the lives of people 900 years ago as a 4K video. The one-minute historical film moves from waking up in a farmhouse to scenes at a market, a blacksmith’s workshop, and inside a castle.
After developing the story and shot list, Umesh used Claude Opus 4.8 to write detailed prompts for each scene, then generated the video with Seedance 2 on BytePlus. The project took a few hours, and the post pulled around 25,000 impressions.
2️⃣ Claude expressed different values by language
ⓒ @claudeai
Anthropic analyzed more than 300,000 anonymized conversations across 11 languages to compare the values Claude expressed. More than 3,000 values were clustered into four axes: deference versus caution, warmth versus rigor, depth versus brevity, and candor versus execution.
Language differences were most pronounced on the warmth-versus-rigor axis. Hindi and Arabic conversations leaned warmer, while Russian conversations leaned more rigorous. Model differences were modest, but Sonnet 4.6 was more playful and affirming, while Opus 4.7 was more likely to offer candid critiques.
3️⃣ Blender camera motion drives Seedance 2
ⓒ @ComfyUI
The ComfyUI team shared a workflow that applies camera motion from a Blender previs to a Seedance 2 video. The final video at the top and the gray previs below follow the same path, pulling back from the players to reveal the full stadium.
The camera’s movement, rotation, and framing are keyframed in Blender, then transferred into Seedance 2. Instead of describing the camera path in text, the workflow demonstrates it with a 3D previs.
That's it for today. We covered AI in a real campaign, historical recreation, language-based AI values, and motion control driven by a 3D camera. We'll be back tomorrow with more. 😄