Release notes

Setup that finishes on slow internet

Installing your AI now copes with slow and patchy connections.

Highlights

• Install Codex or Claude Code on any connection

  You see how much has downloaded, and if your internet drops, it carries on from where it stopped.

Also in this release

• NewReal paintings and archive photos for your videos.
• NewStudio in Vietnamese.
• ImprovedVertical stock video for shorts.
• ImprovedElevenLabs narration on the free plan.
• ImprovedFewer questions before paid work.
• ImprovedBetter guidance for your AI.

Sign in from work, school and VPN networks

Studio now connects the same way your browser does.

Highlights

• Sign in from any network

  If our website opens in your browser, signing in to Studio works too.

Also in this release

• ImprovedBetter guidance for your AI.
• ImprovedClearer reasons when a step fails.
• ImprovedFull quality by default.
• NewA backup for character swaps a video model refuses.

Use Studio from your own AI, even with the window closed

Connect Claude, GitHub Copilot, ChatGPT, Cursor or Google Antigravity in a few steps, with nothing extra to install. Studio can start with Windows and keep running in the background for your AI, and updates install themselves.

Highlights

• Connect your own AI in a few steps

  Choose which AI you use and Studio shows only its steps. Nothing else needs installing. Once connected, your AI walks you through the rest of setup in the conversation.

• Ready in the background

  Studio can start when you sign in to Windows and wait in the system tray, so your AI can use it without the window open. Turn this off in Settings any time.

• Updates install themselves

  When a new version is ready and nothing is running, Studio installs it and reopens in the background. Your AI can tell you an update is ready and install it when you ask.

Also in this release

• FixedCodex setup on protected computers. If your security software briefly holds the new Codex program, Studio waits and tries again. If it stays blocked, Studio tells you how to allow it.
• FixedFirst ChatGPT sign-in for Codex. Signing in now works the first time, without needing a second try.

Put your character into real footage

Give Studio a clip and a photo of your character, and it swaps the person in the clip for your character. Background work on Windows is also quicker to start and to stop.

Highlights

• Swap a person for your character

  Studio replaces one or two people in your clip with the characters from your photos, one photo each. The movement, camera, timing and original sound stay the same.

  Works through your Runware account, at up to 4K.

Also in this release

• ImprovedClearer refusals. When a media service turns a request down, Studio tells you the service’s own reason.
• ImprovedQuicker background work on Windows. Studio no longer starts a helper program before every task, so work starts sooner on a busy computer. Stop and quit close everything Studio started within seconds.

Make videos in the style of one you love

Show Studio a video you like and it works out how it was made. Studio also picks the right way to make each scene, and shows clearly when your AI agent is connected.

Highlights

• Learns from your reference videos

  Give Studio a video you like. It works out the cuts, pace, camera moves, transitions, what is said and the text on screen, then makes something in that style.

  If your own Google, OpenRouter or Runware account offers Gemini Flash, Studio can use it to watch and listen to the clip. Without it, Studio still works the video out on its own.

• The right way to make each scene

  Studio decides per scene whether to use your footage, animation, generated pictures or generated video. It uses a media service only when a scene needs it.

  A service you saved but don’t need for this video no longer holds the production up.

• See at a glance that your agent is connected

  A large green check shows when Claude or Codex is signed in and ready.

Also in this release

• ImprovedSlower slow motion. Clips can now play as slowly as a fifth of their speed, with the sound slowed to match.
• ImprovedBetter vertical videos. Subjects and text are framed for phone screens.
• ImprovedFewer extra steps. Studio no longer adds disclosure requirements of its own to generated media. Each media service’s and platform’s own terms still apply.

Newer models and easier Claude sign-in

The newest Gemini video and GPT image models, vector pictures you can see right in Studio, and a Claude sign-in that no longer gets stuck.

Highlights

• Gemini Omni 1.2

  Make video from text, images or reference clips. Edit a video, extend a clip and upscale it.

  Through Google, Runware, OpenRouter and fal. Runway still offers 1.1.

• GPT Image 2.6

  The newest GPT image model replaces 2.5, and can now make SVG and animated GIF pictures too.

  On Runway, which doesn’t offer 2.6 yet, GPT Image 2.5 replaces GPT Image 2.

• SVG pictures in Studio

  Vector images preview like any other picture, including ones already in your projects. The original file stays downloadable.

• Claude sign-in that always finishes

  If the browser doesn’t open or can’t hand back to Studio, open the sign-in page yourself and paste the code Claude shows you.

Also in this release

• ImprovedClaude plan shown up front. Studio says that a Claude account sign-in needs a paid plan (Pro or Max). On the free plan, use an API key.
• ImprovedClearer agent messages. When your agent is signed in but won’t start, Studio says so and tells you what to try, instead of blaming your internet.
• FixedSafer budgets. When a media service hasn’t confirmed what a job cost, that amount stays set aside in your budget until it does.
• ImprovedReliability. When a media service turns a request down, Studio notes the kind of problem, such as a rejected key, a usage limit or billing, so we can fix the common ones first. It follows your sharing settings and never includes your files, keys or content.

Finishing options and clearer connection help

More ways to polish and export a finished video, and connection checks that tell you what actually went wrong.

Highlights

• More finishing options

  New colour and sound finishing controls, and more export formats, including one ready for streaming.

• Clearer connection help

  Checking an agent or provider connection tests the right one and explains the problem: a rejected key, missing permissions, a usage limit or billing.

Also in this release

• ImprovedEasier local setup. When connecting a tool that runs on your computer, Studio lists every setup problem at once.
• ImprovedReliability. Studio notices more kinds of problems, such as failed connections, saves, imports and recoveries, so we can fix them faster. It follows your sharing settings and never includes your files, keys or content.

Sign-in and discovery improvements

Steadier sign-in and better suggestions for the right tool and provider.

• FixedMore reliable sign-in. Better recovery from saved-session problems and from repeated attempts to return to the app after signing in.
• ImprovedWord-timed transcription. Easier to find when you’re working with an existing voiceover.
• ImprovedClearer suggestions. Audio cleanup is told apart from speech generation, and voice copying from avatars.
• ImprovedMore provider choices. Alternatives stay available for Grok speech and background removal.
• NewProduct analytics. Support for product-improvement analytics.