Recent product updates

New

• Topic Maps can now group discussions by room, making it easier to see what each location is talking about and where the same topic appears across the world.
• The Observatory can now group topic timelines by room, with expandable topic details, activity totals, audience counts, and clickable time periods.
• New worlds can include a ready-made set of Persona Properties for describing an agent’s background, values, decision-making, and communication style.

Fixed

• Stable profile facts are no longer exposed as mutable action state, so simulation turns cannot accidentally rewrite an agent’s identity.
• Renaming or deleting a property field now keeps existing agent, player, and public-profile values synchronized, while invalid property names and values are rejected clearly.
• Switching quickly between agents no longer flashes the previously selected agent’s portrait while the next profile is loading.
• Multi-agent cast changes are now applied atomically, preventing simultaneous roster edits from dropping earlier additions or removals.

New

• Plots can now follow one shared storyline across the world or run separate story progress in each location, with different starting times for concurrent locations.
• Externally controlled and Unreal simulations can now be opened as a read-only live monitor with the 3D scene, activity, and story progress visible in the browser.
• Block characters now include controls for head size, body width, torso, arms, legs, and limb thickness.

Fixed

• Player input now interrupts pending agent responses more reliably, without allowing stale or duplicate turns to appear afterward.
• Agents now keep the intended person separate from location choices when selecting conversation and movement actions, reducing misdirected replies.
• Movement controlled by an external client now stays synchronized when it completes, fails, or is interrupted, so the next turn resumes correctly.
• Block characters now return to a consistent pose between animations and keep their proportions across full 3D and large-crowd views.
• Sign-in sessions, return links, private simulation attachments, and connected services now have stronger security safeguards.

New

• Plot editing now includes a visual graph for creating branching paths, recurring cycles, return points, and final outcomes.
• You can switch between Graph and Table views while editing a plot, making it easier to shape the flow and review the details.
• Live simulations now let you switch between Linear and Graph views of the current plot.

Fixed

• Live plot graphs no longer briefly disappear when the simulation moves to another step.
• Failed World Designer plot changes are now reported as failures instead of being presented as completed work.
• Existing linear plots continue to display and run as a straightforward sequence.

New

• The login screen now supports Google sign-in, with enterprise SSO available for configured organizations.
• Agents can now read public URLs as source material before creating worlds, agents, plots, storyboards, shots, or media prompts.
• Image generation now includes Recraft V4 still/vector models and GPT Image 2 editing for typography, layout, and precise revisions.

Fixed

• Live simulations hold their chat and 3D state more consistently while agent messages are streaming.
• Accounts that use Google or SSO now show clearer password and account-deletion behavior in Security settings.
• Action, agent, and reference-media setup screens now handle parameters, icons, and long media labels more cleanly.

New

• Schedule editors now support blocked time intervals, making it easier to mark breaks, off-hours, and other unavailable periods.
• World settings now include shared agent model and service-tier defaults.
• Agent settings now include clearer model labels, Auto model selection, and service-tier controls.

Fixed

• Scheduled simulations now communicate paused/waiting states more clearly while the scheduler is waiting.
• Local media thumbnails load faster across project and agent surfaces.
• Public API reads for larger worlds, active simulations, and latest logs are more consistent for client apps.

New

• A new API Docs section now gives developers endpoint references, realtime event details, and copyable curl, Python, and TypeScript examples.
• Agent question jobs can answer pair, location, and time-window questions using world, simulation, and log context.
• Similar-agent recommendations can now be generated and fetched for a world or simulation, with filters for a specific agent.

Fixed

• Public API responses now focus on the documented fields developers need for client apps.
• Simulation log and agent list calls now use clearer pagination limits.
• Question answers handle unsupported requests more consistently.

New

• Shared simulations can now be opened with join links, letting guests bring one of their own agents into a host run through a scoped participant view.
• Schedule-based plots now include room calendars, timezone-aware event editing, a simulation clock, and timeline views for following what is happening now and next.
• New YouTube research actions can transcribe videos, sample frames or clips, and describe visual style so agents can work from pasted video links.

Fixed

• Busy room simulations handle waiting, player interruptions, and multiple ready agents more reliably during live runs.
• Shared chats and simulation views now preserve generated media context more consistently when logs are reopened or shared.
• Narrative and storyboard tools are more accurate when placing panels in plot events and attaching shot media assets.

New

• New project workflow starters can prefill behavior sandbox, narrative/story design, virtual event twin, and game design worlds with useful starting context.
• Topic analysis now separates stable and emerging topics, adds semantic topic links, and surfaces agent communities based on shared topic membership.
• Actions can now be assigned to many agents from one modal, and new worlds get a default movement action with location-aware schema setup.

Fixed

• Batch agent generation is more tolerant of duplicate names, already-existing names, and model responses that include trailing content after valid JSON.
• Topic graph layouts scale better for larger semantic networks, with clearer labels, agent overlays, and topic inspector metadata.
• Admins can now reset a user's billing tier override and see the fallback plan that will apply afterward.

New

• Live simulations now include Agent and Meaning graph views, so you can switch between 3D, social structure, and topic-level conversation views while a run is active.
• The simulation canvas now supports fullscreen viewing with clearer controls, current speaker cues, activity badges, and clickable details for agents, communities, and topics.
• Agent creation now includes a Batch mode for generating a cast from one concept, assigning shared actions, and choosing selected or random starting locations.

Fixed

• Stopping a simulation now resolves to a clearer stopped state instead of leaving paused or detached states visible in the UI.
• Participant changes, scheduler waiting messages, and typing state now update more consistently while a simulation is running.
• Simulation logs, generated media links, and game-agent state updates now stay in sync more reliably during active runs.

New

• The Simulation Launcher now uses one browser simulation path with clearer Start and Resume choices, running-run filters, richer run cards, and selected-run stop controls.
• Simulation IDs are now easier to see and copy from chat and resume flows, making it simpler to reopen a run or hand it off to an external runtime.
• Project deletion now includes a focused dependency preview for the cast, spaces, actions, plots, and simulation history that will be removed.

Fixed

• Reopening a running simulation is less likely to create a parallel chat or 3D run instead of returning to the existing session.
• Stopping, pausing, and reconnecting simulations now update visible run status more reliably across the launcher and chat surfaces.
• Project duplication and deletion flows now give clearer feedback while the workspace refreshes around the changed project.

New

• Shared creator, project, character, and chat pages now produce richer preview cards when you post them in social apps, messages, or other link-friendly surfaces.
• Live simulation logs now make it easier to tell who is actively responding, what is still in progress, and how the conversation is unfolding during a run.
• Reference image workflows now give you a faster way to reuse recent media while working on a project.

Fixed

• In-progress chat updates stay readable more often instead of disappearing or showing incomplete fragments.
• Busy simulation moments with multiple overlapping actions now display more cleanly in live views and logs.
• Shared pages and preview images fail less often across different environments and link previews.

New

• Public project publishing now lets you include a narrative snapshot from the default plot, choose one narrative image or video, and decide which shared chat moment leads the public page.
• Public project pages now show that narrative section alongside clearer creator, cast, and featured-scene storytelling so visitors get context faster.
• Reference image and reference video uploads now create the right narrative assets directly, so source material can enter a project without first attaching it to a storyboard or shot.

Fixed

• Featured moments on public project pages now honor the explicitly selected chat moment instead of drifting to an automatically chosen highlight.
• External Meshy-hosted assets now load more reliably through the media proxy, with better handling for stricter upstream access checks.
• 3D model previews and small-screen message layouts now fail more gracefully, reducing broken previews and overflowed metadata.

New

• Actions now support explicit type and target settings, so you can mark an action as conversation or movement and direct it toward a person or a location when that matters.
• World-building flows now surface these action settings directly while creating actions or assigning them to agents, with inline guidance for what each choice means.
• The simulation launcher now calls out missing setup for embodied runs, including 3D character readiness, location coverage, and the movement and conversation actions a world still needs.

Fixed

• Agent chat streaming is less likely to show duplicate or mismatched in-progress messages when tool calls and final logs arrive out of order.
• In-progress tool calls now render more quietly, reducing visual clutter while an agent is still thinking or acting.
• Movement-driven animations in the embodied view resolve more consistently when an action is marked as movement instead of relying only on legacy action names.

New

• Agents can now have a dedicated 3D character workflow with text-to-3D and image-to-3D generation, uploaded models, version history, and preview-ready rigged assets.
• 3D characters can generate animation sets automatically from agent actions or use manually selected motions, with action-to-animation mapping for runtime behavior.
• Public worlds now surface featured moments, creator attribution, and cast previews so it is easier to tell what a world is about before you open it.

Fixed

• Agent chat session state is more resilient during restarts and rapid view changes, reducing cases where stale state could interfere with a run.
• Public world highlights now prioritize the selected featured moment more reliably instead of falling back to a less relevant line.
• Simulation and WebGL loading, controls, and resume behavior are more consistent across the latest workspace flows.

New

• Settings now include Telegram setup for direct chats, with connect links, status, and test messages.
• Telegram replies and quick-action buttons can now route back into the relevant simulation so conversations can continue outside the browser.
• World actions can now send Telegram messages to the linked user when a world needs to notify someone directly.

Fixed

• Agent chat startup is more reliable when the scheduler agent already exists, reducing setup failures in some worlds.
• Room placement and simulation loading behave more consistently during agent-chat initialization and resume flows.

New

• World Designer and Generative Media Artist now ship with default catalog artwork, making curated assistants feel more intentional from the start.
• Simulation launch now offers more guided entry points, including clearer paths for world setup, character imports, and creative exploration.
• Narrative and reference-image workflows now share reusable media grids and schema-driven advanced fal.ai parameter controls.

Fixed

• Plot lookups now behave correctly when a plot identifier is `0`.
• Tagged catalog agents keep their intended default artwork instead of falling back to generic avatars.
• Legacy generic chat simulation titles are recognized more reliably in the simulation flow.

New

• Media creation now gives you more options for quick drafts, polished images, guided edits, and motion work.
• Storyboard, reference, shot, and finishing workflows now surface choices that better match the kind of creative work you are doing.
• Commercial-use information is easier to spot before you commit to an image or video run.

Fixed

• New media options now appear more consistently across the app.
• Usage labels and recommendations behave more predictably across generate, edit, motion, and reference workflows.

New

• New account verification flow with resend support, plus easier password reset and recovery.
• Security settings now include active session management, password changes, and self-serve account deletion.
• Billing now supports generation credit top-ups with a clearer view of balances and usage.

Fixed

• Resolved several signup, verification, and credential handling regressions.
• Fixed generation and deletion issues that could interrupt media and content workflows.
• Creative tools and integrations now behave more reliably in edge cases.

New

• Narrative Designer now supports deeper plot, shot, and storyboard planning workflows.
• Reference images and media links are easier to attach to narrative work as you iterate.

Fixed

• Ownership and deletion support was tightened up around shared app entities.
• Media creation flows are more reliable from start to finish.