PRO Upscaler en @LeonardoAi está en funcionamiento!
El diseño ha llevado meses y puedo decir con seguridad que es único en su especie. La opción FIX es una genialidad y Creative es deslumbrante.
100 Megapixel baby!
https://t.co/tU4m5OB5VF
Qué loco es Seedance 2.0. No había probado en profundidad aún. La gran calve siempre ha sido al generar con IA tener claro el objetivo a donde quieres llegar y una visión muy muy clara. Ya lo vinimos diciendo @dmiguelm hace tiempo y solo hace falta que más realizadores se animen.
🆕 NEW: HeyGen Avatar IV is here.
Our most advanced AI avatar model yet.
📸 One photo.
📝 One script.
🎧 Just your voice.
Most avatars sync to your words. Avatar IV interprets them.
Built on a diffusion-inspired audio-to-expression engine, it analyzes your vocal tone, rhythm, and emotion — then synthesizes photoreal facial motion with temporal realism.
🎭 Head tilts. Pauses. Cadences. Micro-expressions.
➡️ A single image → a video that feels real, not rendered.
Rolling out to all users now.
👇 See examples and what makes it different
And these are just the styles that came to mind. Just imagine how many other variations exist.
Post your favorite Severance edits below! I'd love to see what you're cooking up.
Animated using Kling AI
This year starts off with a great open video model release from @Nvidia Cosmos, which is now natively supported in ComfyUI! Highlights of Cosmos:
1. Most efficient video VAE: Can process 1280x704 121-frame videos on 12GB VRAM GPUs
2. Non-distilled architecture: Better negative prompt handling and training potential
3. Strong image-to-video capabilities: Works from first/last frame or between two images
4. Guaranteed motion: Always produces movement in 121-frame sequences
Check our blog: https://t.co/EAVeVjz3HC
@EgioVictor Joder cómo os molesta que se genere dinero sin engañar a nadie. No como vosotros que prometéis 50 cosas para conseguir votos, acabáis haciendo 0 y os metéis una fiestita para celebrarlo. 🔥🔥🔥
For all the animation part I used mainly @runwayml Gen 3 Alpha, combined with Photoshop in some images after rescaling to make the model understand the movement better.
Some SFX generated with @elevenlabs
And all edited in Davinci Resolve.
Should the ads industry make its spots with AI to reduce costs?
I spent a few days creating this spec ad for Toyota, a fictitious car that doesn't exist.
In the end, it's always the storytelling and your audiovisual skills that count.
#AI#ia#specad#AIArtCommuity
I have used different generative AI tools throughout the process, as they all have pros and cons:
- @midjourney for most of the images.
- @leonardoai for images with text, more control in img2img and Upscale
- @freepik for extreme creative upscale in wide angle shots.
Coca-Cola just shocked the world with their latest holiday ad.
They used AI in a way no one saw coming.
This move might just change how ads are created forever.
Here’s what they did 👇
Mmmm... Pues si llama "mierda" al dataset, qué serían las millones de imágenes generadas por humanos, y "arte" al output que quieres criticar, a lo mejor no está entendiendo muy bien el proceso.
E incluso dicha bien la frase, no seria una definición correcta de IA generativa.