Evolving tools redefine AI video

Google's Veo 2, Kling 1.6, Pika 2.0 & more

Heather Cooper

Dec 20, 2024

Generated on Magnific AI - Super Real model

Total reading time around 5 minutes.

Welcome to Visually AI!

🔮AI News this Week

AI video continues to surpass expectations

The AI video generation space has evolved dramatically in recent weeks, with several major players introducing groundbreaking tools.

Here's a comprehensive look at the current landscape:

Veo 2

Veo 2 (Google) has emerged as a frontrunner, demonstrating superior prompt adherence and video quality compared to its competitors. While it's currently waitlist-only and lacks image upload capabilities, its output quality sets a high benchmark.

Prompt: A close-up tracking shot of a ballet dancer practicing alone in a sunlit studio. Natural light streams through tall windows, casting dramatic shadows. Camera circles smoothly around her as she performs a perfect pirouette. Shallow depth of field focusing on her serene expression. Shot on 85mm lens with soft, dreamy lighting

Kling AI 1.6

Kling AI Video (v1.6) leads the pack with impressive improvements in natural movement, physics simulation, and prompt interpretation. The latest version showcases enhanced coherence and quality, particularly in physical motion sequences.

Prompt: Tracking shot following a lone figure walking through an empty dockyard at night with harsh industrial lights creating deep shadows. Cold metallic grays and blues.

Pika 2.0

Pika 2.0 marks a significant upgrade from its predecessor, introducing an innovative "Ingredients" feature that allows users to combine uploaded images into cohesive videos with realistic placement. There are new templates to upload a person to replace a character in the scene. The improvement in prompt adherence and overall quality is notable.

My selfie uploaded to a new Pika 2.0 template (unbelievable character consistency, including my glasses!)

Runway Gen-3

Runway's Gen-3 excels in specialized features, offering unique capabilities like video-to-video transformation and canvas extension options. It’s capable of creating a cinematic look.

Prompt + input image: Rear tracking shot: Follow the red supercar as it tears through a tunnel of swirling blue and pink light streams. Epic cinematic quality, ultra smooth coherent movement.

Luma AI Dream Machine

Luma AI Dream Machine combines their Photon image model with video generation, featuring a new Boards system for maintaining consistent visual styles across projects.

Photon image + Dream Machine-generated prompt: Watch as a constellation figure interacts with Van Gogh-inspired starry patterns in a sci-fi corridor, creating a vibrant dance of light.

Hailuo MiniMax

Meanwhile, Hailuo's MiniMax impresses with realistic human movement, improved textures, and consistent video quality.

Midjourney image + no text prompt:

Sora

OpenAI's Sora, while innovative, remains unavailable in several regions (including the EU/UK) and produces somewhat unpredictable results.

Prompt: A wide shot of an ancient stone bridge arching over a misty river in an enchanted forest, with bioluminescent mushrooms growing along its edges. The IMAX camera slowly dollies across the bridge, capturing intricate carvings and glowing fungi. Hyper-detailed realism showcases vines and moss-covered stones as ethereal fog rises from the river

Hunyuan Video

Hunyuan Video by Tencent offers a robust open-source alternative with great prompt adherence, natural motion, and quality, but only accessible through platforms like Replicate and Fal or run locally.

Prompt: High-angle shot from above, showing a helicopter landing on a skyscraper rooftop. Twilight with the city lights starting to come alive. Deep blues and blacks with vibrant city lights.

There are several other video models and platforms, including PixVerse, Haiper, Mochi 1 ( open source ), Higgsfield ReelMagic, LTX Studio, and more.

The landscape continues to evolve, with each platform bringing unique strengths to the table. While accessibility varies, the overall trend points toward more sophisticated, user-friendly tools for video generation.

📸 AI Snapshots

Ideogram's new Batch Generator feature lets you upload a CSV or Excel file with prompts and settings to generate images in bulk. Check out my tutorial on X.

Magnific AI launched a new image model called Super Real that realistic images specially designed for professionals, including those in film, photography, architecture, and interior design.

Higgsfield AI’s new ReelMagic multi-agent AI video platform that uses top AI models to turn story ideas into full ready-to-watch videos in a single workflow. Sign up for the waitlist here.

ChatGPT search is available for all Free users now and OpenAI added maps to ChatGPT mobile so you can search for and ask questions about businesses with up-to-date information.

Google Labs’ Whisk experiment lets you use images or text to visualize ideas. Sign up for the waitlist here.

Midjourney opened Relax mode for all users (including Basic plans) for the rest of the year, with near-zero wait time.

Krea AI introduced custom trainings to use in the Krea Editor. Add real products to images or subjects seamlessly.

🛠️ This Week’s AI Tools

Explorer Living Encyclopedia: Odyssey and co-founder of Pixar's Explorer is an image-to-world model that transforms any image into a realized, detailed 3D world.

Google Gemini 2.0 Flash Thinking Model: Faster and more powerful model available in Gemini Chat, now.

Fal sync-lipsync: Generate realistic lipsync animations from audio using advanced algorithms for high-quality synchronization.

Pinokio 3.0: New update with a native Hugging Face API, customizable UI, Debug Mode, Browser Automation API, and more.

💻 Recently On Visually AI Youtube

Open AI Sora - What can it produce? (Results & First Ad Creation)

World Labs | Image to 3D Worlds

📱 Recently On 𝕏

AI Video Model Comparison: https://x.com/HBCoop_/status/1869460712952631724

Pika 2.0 Updates: https://x.com/HBCoop_/status/1867741324482412828

Google Veo Early Access: https://x.com/HBCoop_/status/1868856235253989401

🌎 AI Developments All Over The Globe

Google Veo 2 and Imagen 3 - Google launched its advanced video and image generation models with enhanced capabilities.
Silicon Angle News

Luma Photon Image Generation - Luma Labs introduced Photon and Photon Flash, revolutionary AI image generation models with unprecedented cost-efficiency.
Luma AI Official Page

NitroFusion AI Model - University of Surrey unveiled an open-source AI model enabling real-time image generation on consumer hardware.
University of Surrey Official News

OpenAI Sora Video Generation - OpenAI released Sora, a text-to-video AI tool capable of generating high-quality videos.
The Verge Coverage

Apple Intelligence Image Features - Apple expanded its AI capabilities with Image Playground and Genmoji for enhanced image generation.
Apple Newsroom

Midjourney Style Reference Codes w/ Examples

Included:

• 33 --sref Codes

• 396 Downloadable Images

• 132 Prompts

DOWNLOAD YOUR GUIDE!

Cinematic AI Prompting: Craft Stunning Visual Stories

Camera Shots: Learn essential framing techniques from extreme wide shots to intimate close-ups, understanding when and how to use each for maximum impact.

Camera Angles: Master perspective and emotion through strategic angle choices, from eye-level to Dutch angles, creating powerful visual narratives.

Advanced Techniques: Explore dynamic movement using dolly shots, pan shots, zoom techniques, and Steadicam-style movements in your AI-generated content.

Visual Storytelling: Craft compelling narratives using professional cinematographic elements including:
- Strategic aspect ratio selection (4:3, 16:9, 2.35:1)
- Color palette manipulation for mood and atmosphere
- Lens choice effects and their emotional impact
- Professional-grade prompt writing techniques