Descriptions:
This weekly AI news roundup from AI Search covers an unusually dense cluster of major open-source releases from the third week of December 2025. The video leads with Tencent’s Hunyuan World 1.5, a real-time interactive 3D world generator that users can navigate with keyboard controls as scenes are rendered on the fly. Benchmark comparisons show it outperforming Matrix Game and Gamecraft on reconstruction quality metrics, while requiring only a 14 GB VRAM GPU — a notable efficiency bar for real-time 3D generation.
Other significant releases include Scale, an open-source character animation tool that outperforms One Animate on full-body movements and works across non-humanoid characters; Alibaba’s Turbo Diffusion for offline video generation in roughly 2 seconds; Stereo Space for converting 2D photos to 3D stereoscopic images; and a new open-source image model from Kling. Google’s Gemini 3 Flash and a new Xiaomi open-source language model — claimed to rival GPT 5.2 and Gemini 3 Pro while outperforming DeepSeek and Kimi K2 — round out the language model news.
Additional tools covered include a real-time talking avatar generator requiring only a transcript and reference photo, and a new open-source 3D world explorer with real-time interaction. All featured models include GitHub repositories with local installation instructions, making this a useful snapshot of the competitive open-source AI landscape entering early 2026.
📺 Source: AI Search · Published December 21, 2025
🏷️ Format: Roundup
