November 2026

Saturday, 14 November 2026

ElevenLabs’ Scribe v3 cuts Spanish errors by 35%, according to the company
ElevenLabs’ new speech-to-text models reduce word errors compared with Scribe v2. The figures are the company’s own.

Unreal Engine 5.9 Preview adds MCP so AI agents can create 3D environments
Epic Games has released a preview version of the engine with AI features, all of them experimental.

Three models draw an animal cell diagram in SVG with no content errors
A teacher asked three advanced models for a diagram of an animal cell. None made content errors, and they put the reliability down to the vector format.
A second video with Fable 5.2 adds an agent that checks the scientific content before generating it
After the photosynthesis animation, the same teacher has made one on cellular respiration with a reviewing agent, and has translated the first into Catalan.
“Polvo eres”, a stop-motion animated short made with generative AI
A member of the group has shared their first short film, a spoof of cute horror made with generative AI.

A single prompt is enough for Fable 5.2 to produce a photosynthesis animation for sixth-form students
With a single prompt, the model produced a square video with a natural voice aimed at sixth-form students. The author considers it good enough for class and speaks of a new way to create materials.

How “Buzz” was shot: Claude, Blender via MCP, a virtual reality headset and Seedance
Its director details the process behind the short film the paper covered yesterday, and warns that it is expensive in credits.

A short film is shot inside virtual reality and finished with AI
Raül Torres shares a preview of “Buzz”, a short film whose shots he directs with a virtual reality headset on sets created in Blender.

Google Flow charges credits per second for Gemini Omni 1.2 Flash videos
Flow’s official help page sets out the cost of generating and editing clips with the model, which can already be chosen in the app, according to comments in the group.

Google unveils Gemini Omni 1.2 Flash, which generates and edits videos of up to 15 seconds
The new version of Google’s video model extends clip length and allows clips to be refined over several turns. The announcement was shared in EduMedia-IAG.

A tool turns a mobile phone and VR headset into a virtual camera for Blender
Raül Torres describes how, with help from Claude, he built a system that records real camera movements in Blender to then guide an AI video generator.

A teacher recommends building infographics in HTML and SVG rather than as generated images
A member of Vibe Coding Educativo argues that, to illustrate concepts, AI should generate SVG shapes and icons in an HTML page, with generated images kept for when they are pedagogically necessary.
A website gathers motion-graphics videos with their prompts and skills for agents
MotionVideo offers AI-made videos accompanied by their prompts, and skills that teach a coding agent how to do motion design.

An AI-made horror web series documents its workflow step by step
“Never Sleep Again”, seven episodes released between 25 and 31 October, explains on its website how it combines Blender, Claude, a video generator and DaVinci Resolve.