Viddi AI Blog
Tips, tutorials, and product updates from the Viddi AI team — covering AI video, image, and music creation.
AI Models2026/07/16
What Is Kimi K3? Moonshot AI's 2.5T Parameter Model Explained (Pre-Release)
Kimi K3 is Moonshot AI's upcoming 2.5T MoE model with 1M-token context and linear attention. Here is everything confirmed, leaked, and still unknown before the official launch.
VAViddi AI
AI News2026/07/15
NVIDIA Cosmos 3 Edge: 4B World Foundation Model for On-Device Robots
NVIDIA's Cosmos 3 Edge brings physical AI reasoning to Jetson Thor edge hardware — 4B parameters for real-time robot perception without cloud round-trips.
VAViddi AI
AI News2026/07/14
Candy AI in 2026: Live Action Video, Story Mode, and What's Next
Candy AI shipped 120-second Live Action video clips, Story Mode with context-aware images, the Lumina-Realism V4 engine, and emotional voice mirroring in 2026. Here is what changed and where the platform is heading.
VAViddi AI
AI Models2026/07/14
Bonsai 27B: The 27B Model That Runs on a Phone at 3.9 GB
Prism ML compressed Qwen3.6 27B from 54 GB to 3.9 GB — a 14.2× reduction that lets a 27B-class model run on an iPhone 17 Pro Max.
VAViddi AI
AI Models2026/07/13
What Is Reve 2.1? The 4K Layout-First AI Image Model Explained
Reve 2.1 is a 4K AI image model with a unique layout-first architecture. Learn how it works, how it compares to GPT Image 2 and Midjourney, and whether it fits your workflow.
VAViddi AI
AI Models2026/07/13
Soofi S 31.6B: Germany's Sovereign AI Model with 1M-Token Context
Germany's Soofi S 30B-A3B combines Mamba-2 efficiency, MoE architecture, and full European infrastructure — the continent's most ambitious open-source foundation model.
VAViddi AI