AI tools for creatives: June 2026

June 2026 marked a month of decisive structural changes across the creative technology landscape. The applications that matter for design, motion, and strategy workflows shifted from basic media generation to advanced ecosystem automation. Adobe extended its agentic architecture across the Creative Cloud, Figma introduced timeline tools to challenge legacy motion applications, and Framer embedded custom agents directly onto the canvas.

At the same time, the open-source community introduced compact local models that run entirely offline, and novel niche applications targeted highly specific studio tasks.

Each month I work through what has actually shipped and filter it for what matters in a creative practice. You can find the May 2026 edition and April 2026 edition in the archive.

Main points

• Figma introduced Motion, Code Layers, and custom Shaders at Config 2026; this expands the canvas to eliminate traditional handoff bottlenecks for design teams.

• Adobe rolled out its Creative Agent across Photoshop, Premiere, and Illustrator; this transforms Firefly from an asset generation tool into an orchestration layer for multi-step production jobs.

• Anthropic and OpenAI both filed confidentially for initial public offerings in June; this marks a significant financial turning point as corporate valuations near the one-trillion-dollar milestone.

• Framer 3.0 launched with canvas-native AI agents and Git-style branching; this allows creators to build and audit responsive web layouts safely within an isolated branch.

• Google expanded its open-weight ecosystem with Gemma 4 12B; this compact model runs on standard laptops and uses an encoder-free setup for raw audio and vision.

• The European Commission opened its Code of Practice for signature on 10 June 2026; this establishes official transparency measures for tracking and watermarking generated media assets.

New this month

These tools or integrations entered the market explicitly during the June cycle:

• Seedance 2.0 Mini – ByteDance's fast, high-volume video composition layer

• Framer Agents with MCP – canvas-native development connection for Cursor and Claude Code

• Claude Fable 5 – Anthropic's multi-tier effort reasoning model, launched 9 June and redeployed 1 July following a brief suspension

• Adobe Creative Agent (public beta) – cross-app automation for Photoshop and Premiere

• Figma Motion (open beta) – timeline animation controls inside native design files

• Gemma 4 12B – Google's encoder-free raw vision and audio local model

• DiffusionGemma – open-license local visual style exploration model

General AI tools

ChatGPT

[Major update] ChatGPT reached a milestone of one billion monthly active users. On 26 June, OpenAI retired GPT-4.5 entirely, moving all active user histories to GPT-5.5. Concurrently, GPT-5.6 entered limited preview as a new model family – Sol, the flagship, alongside Terra and Luna – with stronger capabilities across coding, science, and cybersecurity.

The newly introduced ChatGPT Agent changes how the tool operates. Rather than simply writing copy, the agent can complete multi-step tasks across connected applications and web forms autonomously. For commercial studios, OpenAI also launched a $150 million Partner Network to support custom platform integration.

The default interface now features a simplified model tier system. For creative directors managing deep projects, the Responses API serves as the primary development path. It handles compaction for long-running chat histories server-side, making background workflows far more predictable.

The model picker has changed for business accounts, who can select from:

• instant tier parameters for rapid feedback loops

• medium and high settings for balanced strategy creation

• pro standard or pro extended configurations for deep, token-heavy technical analysis

The application has also rolled out upgraded speech-to-text options, native pronunciation assistance in over 60 languages, and personal finance parsing tools for Plus accounts.

Claude

[Major update] Anthropic made major updates to its collaborative ecosystem in June. The company launched Claude Tag, a feature that brings Claude directly into Slack as a shared team member. This turns the tool into a multiplayer workspace where an entire team can interact with the model simultaneously.

Claude Cowork is also generally available on macOS and Windows. It is a computer-use agent – it works autonomously across your desktop, local files, and open applications to complete tasks end-to-end and return a finished deliverable. For creative teams managing multi-stage client campaigns, it removes the need to manually coordinate between tools.

Claude Fable 5 arrived on 9 June, featuring five explicit effort levels to control reasoning token costs. Because Fable 5 performs exceptionally well on short-fiction benchmarks, it is highly practical for scriptwriting, brand narratives, and long-form strategy documents.

Access was suspended on 12 June – three days after launch – due to US export control restrictions. Anthropic redeployed it globally on 1 July once those controls were lifted. If you tried it early and lost access, it is now fully available across Claude.ai, Claude Code, and Claude Cowork.

Anthropic also finalised several operational shifts mid-month:

• two legacy Claude 4 architectures were retired with no grace period on 15 June

• automated API usage limits were split completely from your interactive web app subscriptions

• the baseline token consumption of Fable 5 increased by roughly 30% compared to older models

Perplexity

[Updated] Comet – Perplexity's research browser received a meaningful update in June. Deep Research features arrived in Computer mode, making it more capable for autonomous multi-step desk research. The mid-month changelog also added faster command shorthand access, centralised enterprise security toggles, and expanded research parameters inside their experimental labs. This remains my preferred starting point for parsing complex creative briefs or scanning competitive market landscapes.

Gemini

[Updated] Google introduced several features in June. Gemini 3.5 Live Translate launched, automatically translating more than 70 languages while preserving natural intonation. Computer use was integrated into Gemini 3.5 Flash, which allows you to build custom agents that act across desktop and browser environments.

Google also released Nano Banana 2 Lite, its fastest and most cost-efficient image generation model, built for high-volume rapid prototyping. Gemini Omni Flash entered public preview separately, designed specifically for dynamic video generation and conversational editing workflows. For creators already in the Google ecosystem, these updates turn the application into a capable production assistant, with Google Flow and NotebookLM serving as the most effective starting environments.

Microsoft Copilot

[Updated] Microsoft 365 Copilot added Claude access for licence holders in June, bringing Anthropic's models directly into the platform. Enterprise creative teams can now use Claude without a separate subscription. Copilot remains most relevant for professionals working within established Microsoft infrastructure, and this integration makes it significantly more useful for collaborative writing and strategy work.

DeepSeek and Grok

[Still relevant] DeepSeek saw no headline product updates in June, but its model economics keep downward pressure on pricing across the industry. It is worth tracking for creative technologists interested in open, cost-efficient, and locally deployable options. Grok had no verified product changes in June and remains useful for real-time social monitoring on the X platform.

Creative and visual tools

Adobe Firefly

[Major update] Adobe introduced its public beta of AI Assistants powered by the Adobe Creative Agent. The tool works as an orchestration layer across Photoshop, Premiere, Illustrator, InDesign, and Frame.io, completing complex production tasks based on natural language descriptions.

The June product release included several automated skills:

• brand kit creation to instantly build logos, colour schemes, and visual styles from text summaries

• short product video features that convert static photos into moving e-commerce clips with custom assets

• quick cut tools to automatically organise video sequences around audio scripts or narration voiceovers

• storyboard transitions that turn a written scene sequence into direct video frames

Adobe also previewed its updated Firefly studio with persistent workspace memory across project sessions via custom Elements and Projects folders, which is currently available through a private waitlist.

Other Creative Cloud upgrades

Beyond Firefly, several application updates rolled out across the design suite:

• Assisted Culling in Lightroom, which automatically flags the sharpest candidate image within a near-duplicate stack

• Photo to Video, powered by Firefly and Google Veo, to transform static product stills into moving B-roll

• an updated on-device Remove Tool in Photoshop that runs a generative model entirely offline

• the Object Matte system in After Effects, replacing the legacy Roto Brush with automated AI selection tools

• Concept to Vector in Illustrator to translate rough hand-drawn sketches into clean vector art

Midjourney

[Updated] Midjourney set V8.1 as its default model on 10 June. A new draft option generates 24 images simultaneously at half the typical token cost, making it significantly faster to construct expansive visual moodboards. Combining this batch process with the random style reference parameter lets you analyse different stylistic directions across a single prompt string.

The update also introduced a preview toggle to evaluate the aesthetics of the upcoming V8.2 model, while background training runs began on V9 to fix lingering body coherence errors. 

In an unexpected hardware move, Midjourney unveiled plans for the Midjourney Scanner, a medical ultrasound utility tied to a physical spa facility planned for San Francisco in 2027. While separate from illustration production, it shows the team's long-term interest in alternative visualisation systems.

Canva

[Updated] Canva completed the global rollout of its AI 2.0 system by mid-June. The platform also introduced an official MCP connector. This utility allows designers working inside external environments like Claude or ChatGPT to search template databases, edit canvas layouts, and update brand kits directly within a chat thread, smoothing out lightweight multi-platform asset management.

Ideogram

[Updated] Ideogram launched an MCP integration in June, connecting its text-accurate image generation directly to Claude and Cursor. For anyone using agent workflows, this removes the need to switch applications during image-heavy tasks. Ideogram remains my top choice for any visual generation where clear, readable typography is essential.

Other design tools

The remaining specialised design layout utilities maintain their positions in the workspace:

• Recraft for precise, controlled commercial vector artwork and geometric illustration

• Krea for real-time visual experimentation and fluid canvas-based conceptual prototyping

• Khroma for generating custom colour palettes based on trained human aesthetic selections

• AutoDraw for converting quick digital drawings into clean, predictable icons

• Fontjoy for balancing typographic contrast to build instant font pairings

Video AI tools

Google Flow

[Major update] Google updated Flow to run on the new Veo 3.1 model. This update provides richer audio, sharper prompt adherence, and improved realism for image-to-video generation. Flow also added Gemini Omni Flash in public preview, a model built for dynamic video workflows, alongside an Android beta. 

New canvas controls allow you to:

• use ingredients to video commands to mix specific asset layers

• apply custom frame constraints to direct camera movement transitions

• extend existing video clips or insert specific objects within an established scene

Additionally, Google launched Flow Music on iOS to bring soundtrack generation into the same product ecosystem.

Runway

[Major update] Runway released Agent 2.0 on 25 June, introducing a conversational generation system that analyses inputs and asks clarifying questions to shape video outputs. This moves the interface away from basic prompting toward collaborative editing. 

The company also opened its London headquarters as a European research hub and hosted its annual AI Festival across New York and Los Angeles, expanding the competition categories to touch design, fashion, and advertising. Their custom MCP tool also links video production directly into external developer systems.

Kling

[Updated] Kling 3.0 settled as a primary choice for motion realism in June following its Turbo variant rollout. Built on a Diffusion Transformer architecture, it combines text, image, and editing tools into one pipeline.

Independent testing highlights high marks for physical motion consistency, supporting key studio features:

• native 4K rendering executed at the base pixel layer rather than relying on upscaling filters

• multi-shot sequencing up to 15 seconds that maintains camera and background details across angles

• native audio compilation that creates dialog track adjustments and environment effects simultaneously

• voice binding parameters to automate lip-sync tracking across five core languages

• elements folders to hold facial and character reference sheets across multiple video rendering loops

Luma and Seedance

[Updated] Luma released Ray 3.2 on 3 June, adding multi-keyframe controls that let you direct individual beats throughout a clip instead of just start and end frames. The model also introduced expressive performance transfer for up to eight faces, gesture tracking, and native 16-bit HDR rendering with EXR export. Luma shifted its commercial structure to the Luma Agents platform, with subscriptions starting around $30 per month.

ByteDance released Seedance 2.0 Mini on 15 June, offering a faster and lighter version of its flagship video model for high-volume workflows. It accepts up to 12 reference files simultaneously across images, video clips, and audio tracks to guide composition. Meanwhile, Pika saw no major updates in June and remains best suited for quick social media clips.

Music and audio AI tools

ElevenLabs

[Updated] ElevenLabs focused on infrastructure in June, adding branch merge previews for agents and expanded workspace authentication options via its latest changelog. An update on 8 June deprecated the older v1 speech models, with complete removal set for 9 July. If you use their API, ensure you have migrated to the current models.

Suno and Udio

[Updated] Suno V5 with Studio remains the core offering, but its recent corporate partnership introduced download caps on some plans. Suno is currently defending an RIAA lawsuit on fair use grounds, with a hearing scheduled for July 2026.

Meanwhile, Udio continues to operate under its licensing agreement with Universal Music Group, which includes per-generation royalty payments. I recommend monitoring these legal cases via the Suno and Udio legal tracker at AI Vortex as they will dictate how professional audio generation is handled.

Hume AI

[Worth watching] Hume AI continues to develop its expressive voice tools, including temperature controls for speech generation and its Empathic Voice Interface. It remains a top choice for teams building interactive brand experiences or character-driven audio products where emotional nuance matters.

UI design AI tools

Figma

[Major update] At Config 2026, Figma launched Figma Motion in open beta, placing an animation timeline directly into the design canvas. This tool allows teams to build transitions, spring animations, and easing curves natively, removing the need to export files to After Effects or Jitter.

The conference also introduced several advanced technical canvas tools:

• Code Layers to clone GitHub repositories directly onto the canvas, turning frontend code into editable design layers

• custom WebGPU Shaders created from plain-language descriptions

• Generative plugins that build custom, reusable workflow utilities inside the file

• explicit FigJam and Slides agents to assist with client presentation setups

Google Stitch and v0

[Updated] Google Stitch was updated to use Gemini 3 in June. The platform remains free, providing monthly generation allowances for code-ready interfaces. It translates sketches and text prompts into working UI screens, exporting directly to frameworks like Tailwind, React, and SwiftUI. The v0 platform had no major releases in June and remains an industry standard for turning design intent into production-ready React code.

No-code builders

Framer

[Major update] Framer 3.0 arrived on 16 June, introducing canvas-native Framer Agents. These agents can generate complete layouts, update CMS structures, and audit websites for broken links or accessibility errors. A Git-style branching system lets teams isolate agent actions in a separate workspace before publishing to a live server. With added MCP support for external environments like Cursor and Claude Code, Framer is now fully integrated into the broader developer ecosystem.

Replit, Cursor, and Lovable

[Updated] Replit added a Microsoft Fabric integration for building data-heavy applications, alongside an automated SEO Agent that identifies discoverability issues. Cursor launched an iOS app in public beta across paid plans, enabling mobile agent management, and previewed its upcoming Git platform at the Compile 2026 event. Meanwhile, Lovable addressed security concerns by launching automatic security scanning at no extra cost.

Local and open-source models

Gemma 4 12B

[New] Google expanded its open-weight lineup with Gemma 4 12B on 3 June 2026. This dense model fits within 16GB of unified memory, meaning it runs comfortably on a standard laptop. It uses no separate visual encoders; vision and raw audio flow directly into the main model backbone, making it a highly capable mid-sized local model with native audio processing.

DiffusionGemma

[New] Launched under an open license on 10 June 2026, DiffusionGemma 26B A4B uses a Mixture-of-Experts setup that activates only four billion parameters during use, allowing it to reach high processing speeds on consumer graphics cards. I find this model highly useful for rapid, uncapped visual style exploration without cloud delays.

Local tools infrastructure

Studio engineering frameworks received performance tune-ups:

• Ollama updated its core engine for better execution on Apple Silicon chips

• LM Studio remains a stable desktop interface for comparing open-source models without command line setups

• Mistral released Mistral Small 3.2 with major improvements to instruction following and document text extraction

• Hugging Face added hardware filters to locate models matching your exact physical machine specifications

Novel and niche creative tools

Rather than chasing generic text-to-image features, several specialised tools gained traction by executing distinct creative tasks unusually well:

• Storique builds complete narrative books illustrated in a distinct watercolour aesthetic

• Higgsfield provides an experimental video platform used by independent animators for surreal motion concepts

• Kittl AI transforms rough sketches into complex vector layouts with texture overlays for print production

• Workbeaver takes multi-step control of desktop and browser environments to automate asset sorting and metadata migration

Research and education utilities

NotebookLM

[Major update] NotebookLM received a significant update on 8 June, moving from a standard document reader to an autonomous research agent powered by Gemini 3.5. It supports per-notebook code execution and can source its own web references during a research session. The studio panel now generates audio summaries in 80 languages, alongside mind maps, infographics, and slide decks. Cross-source reasoning is now the default setting, letting you find points of tension across multiple source materials. Google also reduced the Plus subscription price to $4.99 per month.

Khanmigo and Cursor for students

[Still relevant] Khanmigo remains one of the more carefully designed AI tutoring products available, built to guide self-study and mentor tracking rather than replacing active human thought. Meanwhile, Cursor continues its educational program, providing verified students with a complete year of Pro access to help narrow the distance between interface ideas and actual front-end execution.

Worth checking out

The following integrations are worth keeping on your radar:

• Ideogram MCP – an integration connecting text-accurate image generation directly to Claude and Cursor agent workflows

• Canva AI Connector – a tool that lets creative teams manage and edit canvas assets directly from a chat interface

• Flow Music – Google's iOS utility for generating background tracks within its creative application suite

• Runway MCP – connects external video triggering directly to broader canvas automation chains

June 2026 industry trends

Three distinct patterns emerged across the creative technology landscape this month:

• Consolidation around the canvas: dominant design platforms are building built-in timelines and code layers to keep creators inside a single application, eliminating multi-tool handoffs.

• Sovereignty over data: the rise of highly capable 12B and 26B local models allows small studios to handle client data without relying on third-party cloud APIs.

• The move toward physical outputs: novel tools are succeeding by connecting generative assets directly to physical goods, print layouts, and tangible books.

• The rise of GEO and AEO: with audiences consulting models directly, design studios must structure online content clearly so that generative engines cite their work as a trusted reference.

Ethical AI and sustainability reports

European transparency frameworks

The European Commission opened the official signing process for its Code of Practice on 10 June 2026. This framework establishes concrete measures for the tracking, watermarking, and provenance verification of AI-generated assets. For commercial design studios, it provides a structured path to verify asset history and comply with the active enforcement phase of the EU AI Act. You can read the official summary via the European Commission Code of Practice platform. Creative teams can also download the standardised EU icons for labelling AI-generated content to mark deepfakes or synthetic media clearly.

Environmental audits

A major report from Jisc shed light on the environmental impact of generative systems in creative workflows. The data show high consumption of electricity and water within global data centres. This has led studio teams to audit when they choose large cloud rendering passes versus standard local processing, making carbon efficiency a core part of digital asset management. The detailed framework is available to read in the Jisc AI and environmental sustainability report (PDF, 1.2MB) (opens in a new tab).

Previous editions

• AI Notes – May 2026
• AI Tools for Creatives – April 2026
• AI Tools for Creatives – January to March 2026
• AI Notes – September, October, November 2025
• AI Notes – August 2025
• AI Notes – July 2025
• AI Notes – May 2025

Published 7 July 2026. Researched and written by John Luba. Read the comprehensive trends breakdown in the Adobe Creators' Toolkit Report (opens in a new tab) and their 18 June Firefly announcement (opens in a new tab). If you spot any changes, get in touch.