AN IN-DEPTH REPOSITORY
Every model, paper, media release, agent and tool from the month—in one source-linked archive.
COMPLETE MONTHLY ARCHIVE
Creative production moves into everyday tools: image editing, interactive worlds, open 3D pipelines and Arctic opportunities.
THE COMPLETE MONTHLY INDEX
Every linked paper, product release, research project and article has been checked directly. Cards use the publication's own description, imagery and published date wherever the source supplied them.
The AI Search helps discover the wider field, while every available card opens the original publication. Dates come only from publication metadata or official source APIs—never from a video. A checked date records our review, not the release. Older undated entries show their archive month.
Mistral and Mozilla are bringing open, private and multilingual AI to your web browser
Mistral models now power Firefox’s Smart Window beta in France and North America. The assistant works with browser tabs and search; the announcement emphasizes regional language support and provider commitments to zero data retention.
YuE2: Unifying Symbolic and Audio Music Generation at Frontier Quality
YuE2 generates songs through editable melody-and-chord plans, with released weights, cover generation and score-based editing. Edits render a new recording rather than preserving unchanged audio. Code and weights have different licences; company use requires checking the model terms. Release day unconfirmed; source checked 16 September.
Luote-hubin Createch-käräjät 5.10.
The 5 October gathering at Lapland UAS Showroom in Rovaniemi runs 13:00–17:00 and connects creative technology projects, prototypes and business opportunities. Createch North: AI Leap is a partner. Listing checked 14 September; announcement date not supplied.
Small Teams, Ambitious Projects
Blender set a two-year focus on enabling small independent studios to make long-form projects. Animation, collaboration and production pipelines are priorities, with the OVERGROWN feature film informing development.
Introducing DeepSeek-V4.1-Flash
DeepSeek introduced a mixture-of-experts model with native visual understanding and an asymmetric encoder–decoder architecture. The API release targets faster inference and lower cache requirements; the company reports gains over its previous generation.
The Gemini app is now available for Windows
Google released a Windows desktop app with an Alt + Space shortcut for assistance, drafting and image creation. It brings Gemini into desktop workflows without requiring a separate browser visit.
Build more natural voice experiences with GPT-Live-1 in the API
OpenAI made its full-duplex voice model available to developers. It listens and speaks simultaneously, handles interruptions, supports configurable tone and pace, and can delegate reasoning and tool calls to a separate model.
Introducing the Agents API
A public beta brings the Codex agent harness and managed infrastructure to developers. It supports long-running cloud agents, tool use and coordinated subagents, reducing the infrastructure needed to build multi-step workflows.
ComfyUI v0.35.0
ComfyUI added video trim and crop nodes, 3D mesh import, updated partner models and a memory compiler. Transparency, media handling and memory fixes make this a practical workflow update for artists running complex local graphs.
Google deepens its commitment to Finland with a €13 billion investment in AI infrastructure
Google announced a two-year €13 billion commitment to digital infrastructure, energy and partnerships in Finland, plus local investment in Hamina, Kajaani, Muhos and Vaala. This is an announced investment, not completed capacity.
Introducing v6
Suno introduced three music models: v6, v6-wild and v6-mini. The release adds plain-language section and lyric edits, multi-source mashups, and creation from text, audio, images or video. The company says earlier models will retire during the rollout.
Introducing ChatGPT Images 2.5
OpenAI upgraded reference preservation and iterative image editing, added sketch-led creation and reduced generation latency. The release gives creative teams more control over reusable subjects and production revisions.
Introducing Runway for Premiere Pro & After Effects
Runway panels bring generation, upscaling and Aleph 2 clip restyling into Adobe timelines. The plugins are free to download; generation requires a paid Runway plan and uses existing credits.
Forward Deployed Creatives
Comfy launched an enterprise service that embeds creative specialists with teams to validate ideas, build workflows using their own assets and models, and train staff to maintain them. The emphasis is moving working prototypes into repeatable production.
UniMate: One Unified Model to Animate Diverse Skeletons
A research model generates motion from a rigged 3D asset and a text prompt across different skeletal structures. It also demonstrates motion editing, in-betweening and extension without retraining for each skeleton. Paper first submitted 4 September.
WorldSculpt: Generating Compositional Worlds from Grounded Videos
WorldSculpt reconstructs cluttered scenes as individual object meshes from images with instance masks and 3D boxes. The research also demonstrates converting generated Gaussian-splat worlds into separate editable meshes. Paper first submitted 4 September.
Give Your Coding Agents a Memory You Own
Funes turns existing coding-agent sessions into searchable local memory with provenance. It supports agents including Claude Code and Codex, with optional synchronization to a privately owned Hugging Face dataset.
GPT-6 Astra: A new generation of intelligence
OpenAI introduced a new model for computer use, coding, research and professional work, with a staged rollout and additional safeguards for its advanced cybersecurity capabilities.
NeoMME: an efficient Multimodal-native and Multilingual Encoder
H Company released Apache-2.0 multilingual encoders in 260M and 800M sizes that process text and images together, supporting compact visual search across documents, charts and page layouts.
Open Yap 1K: 1,000 hours of full-duplex natural conversation, free for commercial use
A dual-channel English conversation dataset targets full-duplex voice models. An 8.9-hour sample is CC-BY-4.0; the 1,000-hour corpus is available free on request under a separate data-use agreement.
Playco cut manual fixes 50% prototyping games with GPT-6 Astra
Playco connected its AI development environment to Unity and Godot for scene editing, playtesting and iteration. Its case study reports 50% fewer manual fixes and three themed prototypes from one base.
Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States
Researchers jointly model physics, geometry and appearance for 3D world generation and reconstruction. The project releases code, models and a dataset for spatially grounded exploration; it is a research system rather than a finished production tool.
Introducing GWM Worlds 2
A research preview generates interactive 720p video at 24 fps with audio, controlled through camera movement and text actions. It opens possibilities for interactive storytelling and simulation, with remaining consistency and memory limitations.
Sparks Fly: NVIDIA Accelerates Local AI at IFA 2026
NVIDIA outlined easier local-agent setup at IFA, including llama.cpp-based Windows integrations and Perplexity Portable Computer on compatible Linux RTX systems, with Windows support still forthcoming.