New AI waifus, new Deepseek, realtime worlds, Happy Shrimp, tiny TTS: AI NEWS
This update highlights a wave of new open-source AI tools and robotics breakthroughs, including real-time interactive video world generation, 4D human reconstruction from video, compact speech synthesis, improved image and video editors, vision-capable models, and record-setting robots that push speed, jumping, and dexterity.
MAIN POINTS FROM TRANSCRIPT
- Evoke generates interactive, real-time video worlds from images and joystick input, with text prompts adding events and effects.
- 4D Anyone reconstructs a moving person from one video into a detailed 4D Gaussian splat viewable from any angle.
- New open-source tools include an image generator/editor, a tiny text-to-speech model, and ByteDance’s latest video editor.
- Robotics advances include a transformer robot, a tennis-playing robot, and machines setting new sprint and jump records.
TAKEAWAYS
- Open-source AI is rapidly expanding from static generation into controllable, interactive simulations.
- Synthetic world generation could become valuable training data for robots in hospitals, rescue, and exploration.
- Video-based 4D reconstruction is improving enough to produce more consistent, detailed moving character models.
- Robotics progress is accelerating alongside AI, with both performance and versatility improving quickly.