Weekly AI news · 5–11 November · Product news & updates
Weekly AI news: Firefly tools, ElevenLabs Scribe v3 and Gemini Carbon — 12 November 2026
AI news for 5–11 November: Firefly editing tools, ElevenLabs transcription and SoundCloud, Suno challenges, Gemini Carbon, Claude 5.2 and agent updates.
Publisher disclosure: MUVO offers competing music, remix, and video tools. This report summarizes public sources and provider documentation; it is not an independent product test or a claim that MUVO is free of similar issues. Read our reporting method.
This week: editing, speech and agents move into more practical workflows
Adobe added Firefly editing tools, ElevenLabs released Scribe v3 transcription, and Google made Gemini 4 Carbon generally available. Suno added mobile song challenges, ElevenMusic connected to SoundCloud, and Anthropic introduced Fable 5.2 alongside restricted Mythos 5.2 access. Several other announcements concern previews or future changes rather than tools everyone can use today.
This edition covers exactly 5–11 November 2026: 5 November at 00:00 through 12 November at 00:00 Europe/Istanbul (UTC+3). Original sources were rechecked through 12 November at 12:20 Istanbul. Dates below identify announcements or releases, not the time every account received access. Performance claims belong to the providers; MUVO did not run comparative benchmarks or paid generations for this report.
| Development | Date | Availability distinction |
|---|---|---|
| Firefly editing and creative workflows | 10 Nov | Selected tools generally available; Design, Graph and Creative Agent beta; Builder private beta |
| ElevenLabs Scribe v3 / v3 Live | 10 Nov | Recorded and real-time transcription APIs; regional enterprise differences |
| ElevenMusic → SoundCloud | 6 Nov | Linked-account publishing workflow |
| Suno Daily Song Challenge | 6 Nov | Latest iOS/Android app, all plans |
| Gemini Omni 1.2 Flash | 5 Nov | Generally available Gemini API model |
| Gemini 4 Carbon | 10 Nov | Paid API and specified Ultra/enterprise surfaces |
| Claude Fable 5.2 / Mythos 5.2 | 9 Nov | Fable broadly available; Mythos restricted to verified programs |
| ChatGPT voice transition | 10 Nov notice | Retirement planned for 11 December, with exceptions |
Adobe Firefly adds editing tools, with several larger workflows still in beta
Adobe’s 10 November release makes Edit Text, Retouch, Relight and Quick Actions generally available in Firefly. Its broader expansion includes Design, Graph and Creative Agent in beta, while Builder is a private beta with a waitlist. Those labels matter: the announcement is not a universal release of every demonstrated feature. Adobe’s Firefly announcement.
Design brings several kinds of creative work into one canvas. Graph is aimed at repeatable workflows; Creative Agent assists with tasks and supports a planning/approval step. Firefly’s video workflow also gains beta editing features such as timeline work, captions, transitions and B-roll. For creators, the practical question is which parts of an existing editing process can move into the tool without losing control over the result.
Adobe lists newly integrated partner models including Seedance 2.5, Kling 4 and OpenAI image variants. An Adobe integration date is not the upstream model’s launch date, and access or credits can differ by plan.
Creative Cloud assistants and future local integrations
Adobe also announced beta AI assistants across several Creative Cloud products, including Premiere, Illustrator, InDesign, After Effects and Frame.io; Photoshop’s assistant is in its beta application. Its local MCP integration is previewed for a private beta in early 2027, not available today as a completed general release. Creative Cloud announcement.
ElevenLabs Scribe v3 separates recorded transcription from live conversation
ElevenLabs released Scribe v3 and Scribe v3 Live on 10 November. The former handles recorded audio; the latter targets real-time transcription and conversation workflows, including speaker diarization, turn prediction and key terms. The API model identifiers are scribe_v3 and scribe_v3_live. Scribe v3 announcement.
Scribe v3 Live is also available in ElevenAgents. The default service region is the United States; isolated EU, India and Singapore deployments are enterprise arrangements. Creators working with interviews, dialogue or captioning should distinguish a batch transcript from a live interaction system rather than treating both model names as interchangeable.
ElevenLabs claims improved recognition accuracy; those results are its measurements, not MUVO tests. Availability is also distinct from reliability: the separately documented Scribe v3 Live recovery belongs in our incident coverage.
ElevenAI early access and Interaction Engine previews
At its 11 November summit, ElevenLabs announced ElevenAI, with free early access opening that day. The enterprise assistant includes dictation and meeting-note workflows using Scribe v3. It also previewed an Interaction Engine for conversational behavior and routing. Memory, Routines and forthcoming fine-tuning capabilities should be read as announced or upcoming features, not assumed to be generally available. Summit announcement.
Music creation: publish to SoundCloud and try Suno’s mobile challenges
ElevenMusic’s 6 November SoundCloud integration lets creators connect accounts and publish an original track from the music workflow. Linked Accounts is in Settings; the publishing flow includes a description, genre and tags before saving to SoundCloud. This can reduce the handoff between creating a track and sharing it. It does not itself establish broader licensing rights for every input or output. ElevenLabs’ SoundCloud announcement.
Suno introduced Daily Song Challenge on 6 November for all plans in the latest iOS and Android apps. The homepage supplies the daily prompt and a calendar is available in the Library. The release note does not announce a matching web feature. For someone seeking a regular creative exercise, this is a guided starting point rather than a new model or an entitlement to unlimited generation. Suno release notes.
Creator takeaway: choose a small musical brief before generating, then review the result and its rights before publishing. Our song-creation guide provides a practical starting workflow; a publishing integration does not replace those decisions.
Google Gemini Omni 1.2 Flash reaches general availability
Google’s 5 November API changelog makes Gemini Omni 1.2 Flash generally available. The documented video workflow supports generation up to 15 seconds, conversational edits and native audio. Its reference and extension controls have their own limits; Google documents up to six references and extensions of 3–10 seconds per turn, with a total extension ceiling of 30 seconds. Gemini API changelog, model documentation.
The stable model is gemini-omni-1.2-flash. Google also announced increased paid-tier limits and a 3 December routing change for the older 1.1 version. Developers should check their model identifiers and quotas before that date. Model capabilities and hosting-provider limits are separate: a third-party integration may expose a different subset.
Google Vids began rolling out Omni 1.2 support on 5 November, including clips up to 15 seconds and a duration control. The announcement lists personal accounts and specified Workspace/Google AI plans; usage allowances vary. This is an application rollout, separate from the API release. Google Workspace’s November announcements.
For a creator planning a sequence, shot length and editability matter more than treating a longer output as automatically better. See our AI music-video guide for planning shots around a song.
Gemini 4 Carbon becomes available through paid API and Ultra surfaces
Google made Gemini 4 Carbon generally available on 10 November. The announcement identifies the paid Gemini API and Ultra access in the Gemini app, AI Mode, Antigravity and AI Studio, alongside separate enterprise and specialist access arrangements. It is a multimodal-input model with text output; this announcement does not make it an audio or video generator. Google’s announcement, Carbon documentation.
The practical use is reasoning over material and coordinating work, subject to the capabilities of the specific application. Google’s performance and security comparisons remain provider claims. Earlier restricted access should not be confused with this week’s general-availability change.
A separate migration took effect on 6 November: Google’s Gemini Embedding 2 preview identifier now routes to the generally available version. The changelog says the embedding format and pricing remain compatible and that this migration does not require re-indexing existing embeddings. Developers should still update the model identifier in their configuration. API migration notes.
Claude Fable 5.2 is broadly available; Mythos 5.2 remains restricted
Anthropic introduced Fable 5.2 and Mythos 5.2 on 9 November. Fable 5.2 is available through Claude, the Claude Platform and named cloud partners. Mythos 5.2 access remains limited to verified Cyber and Life Sciences programs. The two models therefore do not have the same eligibility. Anthropic’s announcement.
Anthropic lists Fable’s standard API rates at $8 per million input tokens and $40 per million output tokens, with a higher-priced fast mode on specified supported surfaces. Cloud, plan and fast-mode availability must be checked separately. Anthropic’s benchmark comparisons are not MUVO’s measurements.
Claude account and enterprise changes have narrower audiences
Claude’s 5 November personal-finance beta is limited to Pro, Max and Ultra users in the United States. It supports working with linked financial information; this report is describing product access, not recommending financial decisions. Enterprise updates on 5–6 November include security and administration capabilities, several in beta, plus eligible project cost-code tagging. Claude release notes.
OpenAI: restricted security access and a future ChatGPT voice transition
OpenAI’s 5 November GPT-6-Cyber release is restricted to approved organizations through Daybreak Red. Access through supported API and Codex workflows is intended for authorized security work. It is not an unrestricted model launch to every ChatGPT account. Original announcement, deployment safety card.
On 10 November, OpenAI announced that Advanced Voice and Standard Voice would retire on 11 December in favor of GPT-Live. The notes preserve an exception: live video and screen sharing stay on Advanced Voice until GPT-Live supports those capabilities. The retirement is a future commitment, not an event already completed this week. ChatGPT release notes.
ChatGPT Business Hub adds a verified business-management workflow
OpenAI’s 10 November notes introduce Business Hub. Businesses can claim a profile; approval and identity verification are required before publishing information, connecting a product feed or using the specified insights. Eligible U.S. ecommerce and online-service businesses with a website can create an agent, subject to review before publication. These tools do not promise search ranking, citations or traffic. Business Hub documentation.
Coding agents: Claude Code and Codex gain workflow and reliability changes
Claude Code’s releases during 6–10 November add MCP background-task support, configuration controls and model-routing updates. The notes also include permission and redirect-related fixes. Teams should review the actual version’s changelog rather than assuming a model announcement automatically updates an installed CLI. Official Claude Code changelog.
Codex CLI 0.167.0, released on 6 November, adds an agent command center and workflow improvements including compatible-terminal image previews. The 11 November 0.167.2 update enables incremental tool updates by default for Google Cloud Vertex AI, with an opt-out. These are coding-tool changes, not a new creative generation model. Codex changelog.
More creative APIs: Runware integrations and Foundry video previews
Runware’s 5 November entries add hosted access to Gemini Omni 1.2, GPT-image-2.6 Sunburst/Flare and P-Video 2 Pro Replace. On 10 November, it added MiniMax H3 Image variants for generation and multi-reference editing, with advertised support up to 20 megapixels and 16 reference images. Those are Runware’s documented integration capabilities, not a statement that every upstream model launched on those dates. Runware changelog.
Runware’s H3 Image prompt optimizer is enabled by default. Its documentation imposes additional requirements when disabling it, including dimensions and restrictions on reference inputs. Developers should test their request shape against those documented rules before migrating a workflow. No new licensing promise is inferred from model availability.
Microsoft’s Foundry/fal announcement describes a public preview of hosted video models, including Kling v3 Pro in the U.S. Data Zone and LTX 2.5 and MiniMax H3 variants under the listed Global Standard availability. Region and deployment type differ by model. Foundry/fal announcement.
Meta Muse, Mistral Studio and Fish Audio expand assistant workflows
Meta’s 5 November Muse voice and expressive-avatar rollout starts in the United States, across the listed app surfaces. Glasses support is described as coming later. Its 6 November Spaces announcement adds prompt-created interactive workspaces, with boards, tabs and schedules; sharing is still exploratory. Muse voice announcement, Spaces announcement.
Mistral’s 5 November Studio announcement targets enterprise apps, agents and workflows. It combines visual workflow controls with event triggers and deployment choices including hosted, private-cloud and on-premises arrangements. Contact-led enterprise access should not be described as an unrestricted free consumer launch. Mistral Studio.
Fish Audio introduced Live 3 on 10 November for full-duplex voice agents: the system can listen while speaking and handle conversational interruption. English is the supported language at launch. The company lists its agent platform, API and enterprise on-premises options. Latency and quality claims remain Fish Audio’s claims, not a MUVO test. Live 3 announcement.
Workspace features roll out; Microsoft’s EU commitment remains future-facing
Google’s 10 November Gmail announcement expands thread summaries and non-personalized Help me write to additional eligible users aged 18 or over with smart features enabled. The rollout can take up to 15 days and retains user/admin controls. Gmail announcement.
Google Chat’s meeting-scheduling feature starts rolling out from 9 November for specified paid plans, initially in English on the web for direct messages and groups of up to five. A suggested meeting still needs user confirmation. Workspace Studio’s Wait until step supports pauses of up to 14 days inside workflows. These are separate app features with their own eligibility and rollout schedules. Chat scheduling, Studio Wait until.
Microsoft’s 5 November Claude announcement commits to bringing supported Microsoft 365 Copilot use into its EU Data Boundary by the end of 2026. It is a future delivery commitment for the specified EU/EFTA scope, not proof that the transition is already complete. Microsoft announcement.
Research worth separating from an everyday product release
Google’s 10 November AlphaProof Code work explores machine-checked proofs for human-reviewed software specifications, including selected Signal properties. It does not prove every aspect of Signal or every generated program correct. AlphaProof Code research.
Gemini Robotics 2.5, also announced on 10 November, is rolling out to robotics partners. Its industrial examples are evidence for Google’s stated research and partner program, not a consumer creative-app feature or proof of universal real-world reliability. Robotics 2.5 announcement.
What creators should do with this week’s changes
Analysis: the useful pattern is greater control around a generation—editing it, transcribing it, organizing it and publishing it. Pick the workflow that removes a real step from your project, confirm account eligibility, and preserve a review point before sharing output. An announced integration or faster model is not evidence that it fits every production task.
For MUVO readers, the practical next steps are our song-creation guide, stem-separation guide and music-video guide. MUVO’s own app links are provided in the contextual download section below; the products discussed here have separate operators, terms and availability.
Sources
Original announcements, changelogs and documentation are linked beside each story, including Adobe’s Firefly release and ElevenLabs’ Scribe v3 release. We checked the full creative-tool and major-provider watchlist, including Anthropic, OpenAI, Google and ElevenLabs separately. Undated release listings, incomplete client-rendered pages and isolated customer allegations were not turned into launch claims. Older announcements are included only where clearly labeled context supports this week’s change.
This is a selected weekly product-news synthesis, not an exhaustive list of every patch or a claim that every service is healthy. Service failures and recoveries are covered in today’s separate incident roundup. See our reporting method.
Update timeline
- 5–11 November 2026: Seven completed Istanbul calendar days covered by this edition.
- 12 November 2026, 12:20 Istanbul: Original-source verification cutoff; staged rollouts, previews and future dates labeled separately.
- 12 November 2026: First scheduled Thursday edition of the November creator-content experiment prepared for publication.