AI music, voice, images, video & developer updates · Product news & updates

AI news: Suno MIDI uploads, OpenAI Images 2.6 and Luma connectors — 5 November 2026

Suno adds MIDI uploads; OpenAI releases Images 2.6; Luma connects to ChatGPT and Claude. Plus Gemini voice remixing, ElevenLabs integrations and Runware deadlines.

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.

Key product updates at a glance

Suno now documents MIDI uploads for paid desktop-web users, OpenAI has launched Images 2.6, and Luma’s new connector brings its image/video workflows into ChatGPT and Claude. Google’s Gemini API adds voice remixing and broader file handling, while ElevenLabs adds native contact-center integrations. Runware users also have a concrete migration deadline: hosted OpenClaw instances and workspaces are scheduled for deletion on 18 November.

This edition covers 2 November at 11:54 through 5 November at 11:58 Europe/Istanbul (UTC+3). Original announcement dates appear below; this is not a claim to cover the rest of 5 November. Previously reported releases are deduplicated. Service recoveries and customer problems appear separately in today’s incident report.

Product Verified change Availability or limitation
Suno Create MIDI files can guide song creation Desktop web, Pro and Premier
OpenAI Images 2.6 New image generation/editing models ChatGPT/Codex globally; separate API model options
Luma MCP Image/video workflows inside ChatGPT and Claude Connected Luma account; uses Luma credits
Gemini 3.8 TTS Remix an existing voice with instructions AI Studio and voice API; new stored voice
ElevenAgents Four contact-center integrations Platform-specific setup and enablement
Runware hosted OpenClaw Deprecation and workspace deletion announced Export needed before 18 November

Suno: MIDI uploads arrive in Create for Pro and Premier

Suno’s help article updated on 4 November says MIDI upload is available in Create’s Simple mode on desktop web for Pro and Premier plans. Users can provide a .mid or .midi file and describe the desired genre, instruments or feel. The MIDI provides the musical structure; the text prompt guides its presentation. Official instructions and eligibility.

For a producer with an unfinished composition, this offers a way to start from existing note data rather than a text-only prompt. It does not imply mobile availability, a new licensing promise or exact reproduction of every arrangement detail. The help page’s date is a documented update date; no separate launch time is provided.

OpenAI: Images 2.6, GIF creation and saved image references

OpenAI announced Images 2.6 on 4 November, available globally in ChatGPT and Codex across subscription tiers on desktop, mobile and web. Its API offers GPT-Image-2.6 Sunburst and GPT-Image-2.6 Flare; the announcement positions Sunburst for greater detail and consistency and Flare for speed. Those are vendor claims, not MUVO benchmarks. Launch announcement.

The API update includes image markup, SVG export and a complexity control, which is separate from image-quality settings. Developers should check the model and supported output path for their integration rather than assume every consumer feature maps to the API. API changelog.

ChatGPT’s 4 November release notes separately describe GIF creation rolling out, with animation playback-speed controls and sharing. They also add saved reference images that can be reused across chats through an @ label. Model availability and feature rollout are distinct statements; a feature may not appear in every account at the same moment. ChatGPT release notes.

Luma: use image and video workflows inside ChatGPT or Claude

Luma released its MCP connector on 3 November for ChatGPT and Claude. It makes Luma Agent and preset creative workflows available from the assistant conversation, including product shots, alternate camera angles, background changes, relighting and short promotional clips. Results return to the chat. Original announcement.

The connector is tied to the user’s Luma subscription and credits. Connecting an assistant does not make Luma generation free or include it in a ChatGPT or Claude subscription. Luma’s setup page explains the connection flow. Its campaign examples illustrate intended workflows, not measured advertising outcomes from MUVO.

Google Gemini: voice remixing and richer tool-returned media

Remix an existing voice into a new stored voice

Google’s 4 November Gemini API update introduces voice remixing for Gemini 3.8 TTS. A natural-language instruction modifies a prebuilt or designed voice into a new, independent stored voice. It is available interactively in AI Studio or through the voice-creation API using gemini-3.8-flash-tts. The resulting voice can be used with Gemini 3.8 Flash TTS or Flash-Lite TTS. Release notes, voice-remixing guide.

This is useful for adjusting delivery, pitch, energy or timbre without starting a voice design from scratch. It is not a claim that arbitrary third-party voice recordings can be cloned or used without permission.

Office inputs and multimodal function responses expand

The 5 November API notes add legacy PowerPoint, OpenDocument presentations and spreadsheets as inputs, plus page-image processing for legacy Word files. DOC, PPT and ODP are rendered as page images; ODS exposes only the first sheet’s data and formulas. Existing processing for the other listed Office formats is unchanged.

Tools can now return audio, video and Office/OpenDocument files to the model. GenerateContent supports Files API URIs in function responses; Interactions supports document, audio and video result blocks. These are API workflow changes, not a blanket expansion of every Gemini consumer-app upload limit. Dated API details.

ElevenLabs: native integrations for four contact-center platforms

ElevenLabs announced ElevenAgents integrations with NiCE CXone, Genesys Cloud, Amazon Connect and Five9 on 4 November. The aim is to retain existing telephony and routing while connecting AI agents and passing context back to human staff. Official integration announcement.

Setup differs by platform. Genesys uses its native Audio Connector rather than requiring a SIP configuration; Five9 requires its AI Agent Connect add-on and account-team enablement. NiCE setup involves sales, while Amazon Connect has a documented integration path. Other contact centers can still use SIP.

This is an integration release, not a new music model or an automatic entitlement for every ElevenLabs plan. Teams should confirm their contact-center requirements before treating the announcement as immediately usable in an existing deployment.

Runware: hosted OpenClaw deadline and new media-model options

Runware’s 4 November changelog says hosted OpenClaw will shut down on 18 November, deleting every hosted instance and its workspace. Users should copy material they need before that date. The suggested migration is self-hosted OpenClaw using Runware’s compatible APIs and skills. Deprecation and migration notice.

The same changelog documents these operator-specific additions:

These dates establish availability on Runware, not original launches or availability in every upstream provider’s consumer app. The hosting shutdown does not mean Runware’s image, video or audio APIs are being retired. Full dated changelog.

Anthropic: FlashProve explores proof of which model answered

Anthropic published FlashProve on 2 November as an open-source research prototype for provable inference. It aims to let a client verify that a response came from committed model weights without receiving those weights. The release includes code, a technical report and formal proofs. Original research.

The limitations are substantial: the published experiment uses a particular quantized model and inference setup, and proving remains much more expensive than ordinary inference. The repository labels the project an unmaintained research artifact. It is not an announced production feature of Claude or its API. Repository and scope.

Analysis: Model provenance could become useful for customers buying routed AI services, but proof of which model ran would not by itself establish that its answer is correct.

Microsoft’s 4 November release-note batch covers updates released from 28 October through 3 November. The named apps and platforms matter; this is not a single all-users rollout date. Official Microsoft 365 Copilot notes.

These changes help users inspect edits and move work between tools. Licensing, connected data and organization policy still determine practical access.

Other verified workspace, education and developer updates

ChatGPT Dots: enterprise rollout includes an admin preview

The 4 November Enterprise/Edu release notes announce Dots for eligible Enterprise, Edu and Healthcare workspaces with an admin-preview period through 18 November. Administrators can enable it or opt out before the default rollout. The notes exclude UAE data/inference residency; EU and US inference-residency workspaces are supported. Admin controls include connections, activity and pause/resume. Official eligibility and rollout.

Classroom: Guided Learning can be attached to assignments

Google announced the global rollout on 4 November, beginning 3 November and taking longer than 15 days. Teachers can add structured Gemini Guided Learning activities to assignments. Availability covers Education Fundamentals, Standard and Plus, in Gemini-supported languages. This is a staged integration, not proof that every school has it yet. Original Workspace announcement.

OpenAI security research and restricted GovCloud access

Daybreak Red’s GPT-5.6 Cyber is now available to approved customers in AWS GovCloud (US-East), according to OpenAI’s 4 November announcement. Access needs both OpenAI approval and AWS provisioning; one does not automatically enable the other. Availability notice.

Separately, OpenAI reports that a research pipeline validated 948 Linux local privilege-escalation exploits and generated candidate patches. Findings and patches undergo human and maintainer review; this is not a claim that 948 fixes are already installed or that an upcoming model is generally available. Original research and limitations.

OpenAI’s proposed Pierre acquisition concerns Codex tooling

OpenAI announced an agreement to acquire Pierre on 4 November, subject to closing conditions and regulatory approvals. Pierre’s Trees and Diffs libraries are already used in Codex; future Code Storage integration is an intention, not a newly shipped feature. OpenAI says Pierre’s projects will remain open source and Code Storage customers can continue using the service after closing. Announcement.

Sources

Dated original sources are linked beside each claim, including Suno’s MIDI instructions, OpenAI’s Images 2.6 announcement, Luma’s connector announcement, Gemini API release notes and Runware’s changelog. Announcements, staged rollouts, API integrations and research prototypes are labeled separately. We did not run paid generations or benchmark vendor quality and speed claims. The broader music, video and major-provider watchlist was checked; an inaccessible page or lack of a qualifying update does not prove nothing changed.

For the previous week’s context, see our seven-day AI news summary. MUVO publishes these articles and offers competing creative tools. Our reporting method explains the evidence standard and commercial disclosure.

Update timeline