> For the complete documentation index, see [llms.txt](https://docs.atlas.design/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.atlas.design/atlas-ai-studio-overview/release-notes/release-0.39.0.md).
# Release 0.39.0
Previous Version: **v0.38.1**
Save alternative project directions, organize your canvas, and build image and video workflows with new grounding and tracking tools. These are the new user-facing features since 0.38.1. Available nodes, models, and controls depend on your account and workspace access.
## Shared project versions
Save a named version, branch from an earlier direction, and return to it from the new **version tree** in the editor's fourth navigation segment. A version keeps the graph, layout, annotations, selected outputs, and AI conversation together.
When you switch with unsaved edits, choose **Save as new version and switch**, **Discard and switch**, or **Cancel**. The checked-out version is shared: switching changes the project for everyone who has it open. Creating and switching versions requires edit access and waits for active runs, scheduled work, processing, and AI work to finish or be canceled.
Uploaded files and completed run histories stay shared across versions, including runs from deleted nodes. Export uses your current working edits; copying a project starts a separate version tree. This release supports creating and switching versions; merging, renaming, and deleting versions are not included.
{% embed url="" %}
## Browse reusable workflow templates
The new **Templates** page brings reusable workflows together with thumbnails, titles, node counts, and **Text**, **Image**, **3D**, **Video**, and **Audio** hints. Use **Contains** to find a media type; mixed workflows appear under every applicable type.
Choosing a template creates a named copy in your active workspace and opens it for editing. Nothing runs or spends generation credits until you run nodes. The Templates entry appears when a catalog is available, and copied workflows retain your workspace's model permissions.
{% embed url="" %}
## Canvas frames with shared family colors
Groups and sections now share a tinted frame design with an in-frame title, optional description, and a wider drag surface. Nested frames become quieter with depth, while moving a frame carries its contained frames, notes, and owned nodes with it.
Double-click a frame or use its menu to rename it. Choose from six family tints, or **Remove section, keep nodes**. Changing a top-level family's color is one undoable edit. **Ctrl+G** wraps a selection in a section, and the toolbar now offers **Select**, **Pan**, **Section**, and **Note**. Read-only embeds use the same frame appearance.
## Draggable canvas note cards
Notes now use a neutral card with a **NOTE** header. Put a title on the first line and supporting text below it. Click to select, drag from across the card, and double-click or use the pencil to edit. Escape or clicking outside saves your text.
Resizing, body-size controls, undo/redo, and saved placement remain available. Read-only views show the same cards without editing controls.
## See which nodes are running
Running executable nodes now have a subtle pulsing glow that adjusts with canvas zoom. It stays steady when reduced motion is enabled and disappears when the run finishes.
## Nano Banana 2.1 for image generation and editing
**Nano Banana 2.1** becomes the default engine for newly created **Quality Text → Image** and both single-image and image-list **Text + Image → Image** nodes. Choose **Minimal**, **Medium**, or **High** reasoning, with **Medium** selected by default. Existing saved engine choices and paid outputs are preserved.
## Ground results with Google Search and image references
Supported Gemini engines gain an optional **Google Search** toggle in image generation and editing, **Text Generation**, **Structured Output**, and single-image and image-list caption nodes. **Image Search** is available for Nano Banana 2.1 and Gemini 3.1 Flash Image. Both toggles start off.
Search-enabled results **can show** Google's search suggestions, clickable source links, and a compact image-reference grid. Select a reference to open its source webpage. These are result previews, rather than new connectable workflow outputs; their availability depends on what the run returns and the platform's evidence-storage setting. Search-enabled runs execute freshly rather than reusing a shared grounded-result cache.
## Camera-orbit videos and selected viewing angles
**Video Camera Orbit** turns a source image into a camera-motion video using **MiniMax H3 360 Orbit (LoRA)** or **H3 Max Camera Controls**. Choose level, overhead, and low-angle orbits, or movements such as dolly, spiral, crane, and swing. Duration is adjustable from **5 to 15 seconds**.
{% embed url="" %}
**Custom Angle** returns a trajectory video and an approximate selected-angle image. Choose **Tracker + Gemini** or **Gemini only**, optionally provide surface landmarks, and use **Appearance Prompt** to describe hidden surfaces. You can also replace the custom-angle prompt completely. H3 Max uses camera keyframes; LoRA follows editable prompts.
Generated motion and unseen geometry are approximate, and side-angle selection can be unreliable. Video Camera Orbit is unavailable to accounts using the `ip_safe` role.
## More room for precise region editing
Expand the visual mask or region picker into a larger modal for precise work, then press **Escape** to close it. The enlarged editor also supports draggable landmark points and their labels.
## Live rigging-worker availability
Rigging nodes now show whether the shared worker is **available**, **busy**, **unreachable**, or **unavailable**. A busy status can include the operation and elapsed time. This is informational: it does not reserve the worker or disable **Run**. Signed-in API clients can also read the sanitized status at `GET /api/rigging-tools/status`.
## Enable on-demand usage while staying on Free
Eligible workspace owners and admins can add a card through Stripe's **zero-charge checkout** and enable on-demand usage without buying a paid plan. Start from **Plan & credits**, the credit popover, or the blocked-run card.
Before first enabling on-demand, or raising a cap without a current agreement, Atlas shows the terms and asks for explicit agreement. Lowering or disabling a cap does not require a new agreement.
The workspace stays on **Free**, with a selectable cap up to **€100 per period** and no Unlimited option. Turning on-demand off and back on restores the remembered cap. Card-only workspaces are identified as **Free · On-demand**.
## Readable frame titles when zoomed out
Top-level frame titles scale up relative to the graph when you zoom out, making sections easier to identify. The enlarged title lifts above its frame, leaving the contents clear.
Nested frame titles hide below 40% zoom to reduce clutter. Hover, select, or rename a nested frame to bring its title back.
## Return to your template copy
Open a template once to make an editable copy in your active workspace. Its card then shows **Your copy**. Open it again to return to the same project with your edits intact, instead of making another copy.
The saved copy is remembered for your account in that workspace. If you delete it or lose edit access, opening the template makes a fresh copy. Copies created before this feature are not automatically linked to the card.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://docs.atlas.design/atlas-ai-studio-overview/release-notes/release-0.39.0.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
