New AI waifus, new Deepseek, realtime worlds, Happy Shrimp, tiny TTS: AI NEWS
Open-source local models are rapidly advancing into real-time interactive video and direct pixel-space rendering, offering filmmakers granular visual control without subscription paywalls.
What this lesson covers
This news round-up covers newly released open-source AI models across real-time video generation, 4D Gaussian splatting, and native pixel-space image editing. Key releases include Evoke for real-time interactive video, 4DAnyone for turning video into 4D character splats, SenseNova U1.5 for native 4K image generation without VAE decoders, and Bernini v2 for natural language video editing.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 00:57 ↗
Evoke generates 14-billion parameter interactive real-time video responsive to joystick and text inputs in just three diffusion steps.
- 03:05 ↗
4DAnyone extracts a 3D skeleton from a single input video to synthesize new camera angles and reconstruct 4D Gaussian splats.
- 04:28 ↗
SenseNova U1.5 generates native 4K resolution images directly in pixel space, skipping traditional latent VAE encoders and decoders entirely.
- 06:25 ↗
Bernini v2 provides omnimodal, prompt-driven video editing capable of shifting camera perspectives and swapping backgrounds.
Workflow outlined in the video
- Download the Apache 2-licensed Evoke model weights from ModelScope or GitHub to experiment with real-time scene generation locally.
- Run 4DAnyone using a mid-to-high VRAM GPU to extract animated 4D character assets from recorded video footage.
- Test native 4K image editing with SenseNova U1.5 by using direct text annotations on source image canvases.
- Evaluate Bernini v2 workflows in local setups if high VRAM capacity is available for natural language video modification.
Before you use this workflow
These notes describe the source video at its publication date. Model access, pricing, connectors and interfaces may have changed. Check the original source and the provider’s current documentation before installing an add-on, connecting an account or spending credits.
ReelStack has not independently tested this workflow. Preview one representative shot and check motion, continuity and output quality before applying it to a full production. No result, cost saving or model capability is guaranteed.
How to interpret the numbers
135,335 views captured 27 September 2026. ReelStack recommendation score: 38/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 45% performance: 0.88× views vs 7 other channel videos observed at a similar age. View-evidence factor 99% (views / (views + 1,000))
- 25% freshness: 18/100 with a 14-day half-life
- 20% momentum: 174 views/day vs 680 channel baseline (5 comparable uploads), with the same view-evidence factor
- 10% engagement: 82/100; likes + 4× comments, smoothed with a 500-view neutral prior