OpenAI hacked, Jev, Google’s RSI, Qwen 3.8 Omni, Bonsai 2, new Gemini Live: AI NEWS
AI filmmakers can leverage Viggle's Meridian to re-render existing video shots from alternate camera angles using point-cloud depth estimation, giving post-hoc camera direction without full re-prompting.
What this lesson covers
This news breakdown covers major AI updates including Viggle's Meridian for post-generation camera angle control, the lightweight R2T2 transcription model, and X-Gen Labs' persistent world simulation framework Jing and Dao. It also examines Google's Dream RSI recursive self-improvement project and recent model benchmarks.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 01:13 ↗
Viggle's Meridian uses VGGT Omega geometry depth estimation to build a 3D point cloud of an existing video, allowing creators to reshoot the scene along new camera paths with MiniMax H3.
- 02:41 ↗
R2T2 is a compact (~4 GB) open-source model offering low-latency, real-time speech transcription that benchmarks competitively with GPT Live Transcribe.
- 04:12 ↗
X-Gen Labs' Jing and Dao separate world state simulation (Dao) from subjective visual rendering (Jing), enabling persistent multi-agent environments.
- 06:34 ↗
Google's Dream RSI enables AI agents to improve search and problem-solving strategies using discovery trees without modifying base model weights.
Workflow outlined in the video
- Deploy the open-source Meridian codebase locally to adjust camera angles, pans, and orbits on existing video clips.
- Generate 3D point clouds from input footage via VGGT Omega depth estimation to plot custom virtual camera paths.
- Pass reconstructed camera trajectory roughs into MiniMax H3 to re-render high-fidelity video angles.
- Run the lightweight R2T2 model locally on consumer GPUs for low-latency live audio transcription.
- Implement persistent multi-perspective scene generation using the Jing and Dao world simulation framework.
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
58,440 views captured 20 September 2026. ReelStack recommendation score: 45/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 65% performance: 0.24× current views vs the channel's recent median; same-age history is not yet available. View-evidence factor 98% (views / (views + 1,000))
- 25% freshness: 99/100 with a 14-day half-life
- Momentum pending: collecting daily snapshots; its 20% weight goes to observed performance, not free points
- 10% engagement: 77/100; likes + 4× comments, smoothed with a 500-view neutral prior