Release - March 16, 2026

Audio and video transformation nodes in Blueprints: Eight new utility nodes are now available in the workflows node-based editor: Trim Video, Trim Audio, Resize Video, Split Video & Audio, Hue & Saturation, Extract Frame (with timestamp support), and a Reference Audio node for attaching audio references to generation tasks. A further five nodes ship today adding Volume Control, Speed Adjustment, and Multi-track Mixing, rounding out the core audio/video processing toolkit.

Forge support in MCP App: The MCP interface now supports both workflow and forge generation modes, with auto-detection and real-time progress tracking. Previously limited to workflows, agents can now run forge generations and monitor their output directly from the MCP app UI. In order to access it, add https://mcp.app.layer.ai/mcp as a connector to your Claude.

Text node with runtime input support: A new Text node handles string and string list values with runtime input support and conversion to and from prompt format, enabling more flexible text handling within blueprint workflows.

New AI image models: Six new image generation models are now available. Qwen Image 2 (standard and pro variants) supports text-to-image and image editing. Grok Imagine Image supports both generation and editing, expanding your options for high-quality image output.

Social media size presets: Instagram, Facebook, LinkedIn, and other platform-optimized size presets are now available in image composition workflows, removing the need to manually enter dimensions for common formats.

MCP now open to all authenticated users: MCP server access is no longer restricted to internal users and is available to all authenticated accounts.

Multi-workspace support & improvements: Layer now fully support parent/child workspace structure that lets studios manage Creative Unit (CU) budgets across multiple teams or sub-studios from a single place. Check out this quick guide for more info.

Prompt Template node in Wokrflows: A new Prompt Template node lets you create reusable prompt templates using @mention syntax, which converts to dynamic placeholders at runtime. Useful for teams standardizing prompt structures across multiple workflows. GA coming soon, ask for access!

New aspect ratios for Nano Banana: 4:1, 8:1, 1:4, 1:8, you can find them under ULTRA section.

Activity feed access restored for enterprise customers: Fixed an incorrect permission check that was preventing enterprise users from accessing the activity feed.

File picker thumbnail loading: The file picker now loads thumbnails instead of full-size images, preventing several hundred MBs from loading unnecessarily when browsing recent files.

Style defaults no longer overwrite session values: Fixed a bug where reopening a previously saved inference form would reset user-configured values back to style defaults.

Form validation on untouched fields: Validation no longer triggers on fields the user has not interacted with when other fields become disabled, reducing misleading error states.

Video creation from object tiles: Right-clicking an object to create a video now correctly extracts the first frame instead of passing an invalid array.

Reference image dimension snapping: The dimensions field now automatically adjusts when a reference image changes, including for models that do not provide size presets.

Five production frontend errors resolved: Fixes address crashes affecting 30+ users, including issues in workflows version mapping and file handling.

File upload encoding fixed: Files uploaded from URLs are now stored with correct encoding metadata, preventing corrupted downloads.

LTX v2.3 video generation restored: Updated the model endpoint to resolve failures that were preventing video generation from completing. (shipped March 9)

Workspace creation redirect fixed: Users are now properly navigated into their new workspace after creation instead of being left on the creation page.