Product Changelog
For writers, creators, filmmakers, and early customers.
Since v3.0, Virgil has grown from a desktop screenwriting assistant into a full AI creative workspace for writers and filmmakers, now in public beta with a 14-day free trial.
It now brings writing, storyboards, video editing, original character voices, music, and sound into one creative workspace, alongside multi-provider chat, Agent Mode, and research with sources.
Behind the scenes, Virgil has matured with stronger authentication, Terms of Service enforcement, tenant isolation, secure media delivery, model fallback infrastructure, pricing manifests, desktop auto-updates, and a public-launch security pass.
Unreleased - Conversation Recovery After Time Away
Improved session and connection recovery when returning after time away or waking your computer, across desktop and web.
Catch up on messages and replies already saved while the connection was interrupted, without sending your request again.
Keep replies from briefly appearing twice as they finish, with a quiet connection notice that clears when recovery succeeds.
Keep your unsent draft, attachments, and editor work in place during recovery, with a clear sign-in prompt when automatic recovery cannot finish.
A request whose status is still uncertain remains pending until completion or a stop is confirmed. Reconnecting does not resend it.
v3.4.2 - New Models and More Reliable Creative Workflows
More choices for writing, images, video, and voice, with fixes that help conversations, storyboard work, and usage records stay in sync.
Updated the OpenAI default to GPT-6.1 Sol, with Maximum reasoning available for demanding creative work. Specific model selection keeps GPT-6 Astra, GPT-6 Luna, and other supported OpenAI models within reach.
Updated the Anthropic and xAI defaults to Claude Opus 5.5 and Grok 4.7. Claude Sonnet 5.5 and Fable 5.1 remain available when you choose a specific model.
Added image choices including FLUX 3 Image, Ideogram 4.5 Precise Edit, and Seedream 5.0 Flash. Nano Banana 2.1 is available for manual selection, while existing image options remain available.
Added Gemini 3.8 Flash and Flash-Lite TTS, plus Eleven v4 and v4 Turbo, with saved voice takes and improved playback.
Fixed large-storyboard reads and missing final answers in Agent Mode. Replies now persist through reloads, with duplicate assistant delivery suppressed.
Improved image-generation and edit usage records, including Nano Banana 2.1, so model costs and usage are recorded consistently without duplicate entries.
Improved handling of generated takes and timing. Original media stays available when you revise a shot; caption and automatic assembly controls require usable timing data.
Improved saved generation status, prompts, and model details so you can review how an image or clip was created.
v3.4.1 - Video Studio, Character Voices, and Sound
Bring your storyboards, characters, dialogue, and music into a simpler place to build a scene. Ask Virgil for help or choose the creative tools yourself, with your video and timeline kept in view.
Redesigned Video Studio with a warm, dark palette and one collapsible, resizable left panel. Switch between Virgil, Video, Characters, Music, Sound, Storyboard, References, and Media without covering your cut.
Generate a clip with a prompt, model, starting image, and simple settings. Model menus now include provider logos and guidance, and new manual video requests start with Seedance when available.
Bring saved storyboard images into video generation as starting frames, or add storyboard and media assets to the timeline with drag and drop or an Add button. Generated images, video, and audio stay together in project history.
Design an original character voice with ElevenLabs: describe how it should sound, listen to the auditions, and save your favorite for later dialogue and character casting.
Keep approved character appearance references and voice choices together in reusable, versioned sets. New supported generations can use those references while existing shots keep the version you selected.
Create voice recordings, dialogue and table reads, music, sound effects, and background ambience. Preview the resulting takes from the media library and add the ones you want to your scene.
Animate a portrait with an existing recording, or use a recording to lip-sync a video. These tools work with selected excerpts and preserve your original audio in the new take; review the result before using it in your cut.
Edit picture, titles, and sound on the timeline, with trimming, splitting, track controls, and Undo. Play handles preview preparation, and Export creates a video from the saved cut.
Find clip actions with a right-click, or use the plus at a selected clip’s end to open Extend. Generating a new continuation is experimental and depends on the source and supported model; review it before adding it to the timeline.
Export ordered storyboard pages as a PDF and an editable treatment package for sharing a pitch. Sora has been retired from new video generation.
v3.4.0 - A Refined Workspace and More Reliable Writing Tools
More room for your writing, clearer conversations, and smoother movement from an idea in chat to a proposal in your screenplay.
Consolidated the project title and workspace tabs into a compact desktop header, with softer panel edges and clearer space between the editor and chat.
Refreshed Projects, Settings, and the Style Library. Project search, sorting, and a single scrolling list make it easier to find your work.
Brought conversation titles and tabs together, improved title persistence, fixed outline ruler scrolling, and removed the length limit on custom AI instructions.
Chat now uses AI to recognize when a request needs Agent Mode, including follow-up instructions, while preserving your selected model. Unclear requests get a choice before switching.
Fixed duplicate creative replies and working indicators that could remain after a task finished. Working cards now stay out of the way while responses stream.
Fixed insertion proposals that could leave an empty screenplay blank, and improved the handling of accepted edits, review completion, and Undo.
Fixed imported video duration so clips retain their correct length in the library and timeline.
Polished the guided demo with clearer scene formatting, smoother navigation, and a working copy-and-paste flow into the screenplay editor.
Updated the Fireworks-backed DeepSeek and Qwen options to DeepSeek V4 Pro 0813 and Qwen 3.8 Max, and improved chat completion, connection recovery, and cancellation handling.
Upgraded the desktop runtime to Electron 44. This version requires macOS 13 or newer on Apple silicon; Windows remains supported.
v3.3.6 - In-App Support, Agent Media Fixes, And New Image Models
Added a Support tab in Settings so you can report bugs or reach the team without leaving the app. Replies go straight to your email.
Fixed agent-driven image and video generation issues so media requests made through Agent Mode complete reliably.
Added Recraft 4.1 image models and MiniMax H3 Max video generation.
Refreshed the model catalog and tightened image model pricing checks.
v3.3.5 - Document Upload Improvements
Improved document uploads across the app, making it smoother to bring your own manuscripts and pages into Virgil.
v3.3.4 - Faster, More Reliable Releases
Optimized the release pipeline so updates build and ship faster on both platforms.
v3.3.3 - Update Flow Fixes
Fixed issues in the desktop auto-update flow so new versions install more smoothly.
v3.3.2 - Usage Dashboard Upgrade
Upgraded the in-app usage dashboard with model rankings, most-used and highest-cost insights, and clearer estimates of what your usage is worth.
Improved trial and complimentary billing states so what you owe (or do not owe) is always explicit.
Fixed video thumbnails in the media library.
v3.3.1 - Sign-In Reliability
Fixed a sign-in bug that could prevent conversations from loading after login.
v3.3.0 - Public Beta Launch
Virgil opened its doors. The public beta brings a free trial, transparent usage budgets, and a new website with a hosted live demo.
Launched the public beta with a 14-day free trial including $20 of AI usage, protected by a hard budget cap so trials can never overspend.
Added beta seat reservations and a streamlined checkout for Starter, Pro, and Studio plans.
Connected trial billing end to end with Stripe, including usage metering for tokens, images, and video.
Shipped the new virgil.com.ai with a hosted, guided demo of the real app.
v3.2.9 - Launch Readiness And Native Claude
Moved Claude integration to the native Anthropic SDK for faster, more reliable agent runs.
Improved agent behavior and fixed UI issues in outline editing.
Qualified every image and video model in a full pre-launch audit, with alerting for billing integrity and model routing health.
Introduced model tiering so each plan gets the right mix of speed, quality, and cost.
v3.2.8 - Agent Stability
Fixed agent glitches that could interrupt longer runs.
v3.2.7 - Video Studio Polish
Fixed mute and solo controls on the Video Studio timeline.
Final stability fixes across the editing surfaces.
v3.2.6 - Studio Agent, Video Editor Flow, And A Bigger Model Catalog
A summer of work on making Virgil's agent a real production partner: it can now drive the storyboard and video editor, not just the page.
Taught Agent Mode to operate the storyboard and Video Studio, so one request can flow from script to shots to a cut.
Added the Virgil dock and proposal tray, giving agent suggestions a proper home in the UI.
Added a new model selector with an Auto mode that picks the cheapest model capable of the job and escalates when the work gets hard.
Expanded the model catalog with new video generation models including Seedance, Wan, Kling, and Hailuo.
Improved the hosted demo and web experience.
v3.2.5 - Public Launch Readiness
Virgil is being hardened for a larger public audience. This release focuses on trust, reliability, safer media handling, and keeping the app current as AI models evolve.
Added a Terms of Service acceptance flow directly inside the app.
Strengthened account, project, conversation, media, and editor protections across the product.
Made storyboard and character images more secure by moving protected media away from token-in-URL loading.
Added a model registry so Virgil can keep pace with new OpenAI, Claude, Gemini, Grok, OSS, DeepSeek, Qwen, Kimi, video, and image models.
Added smarter model fallback and pricing infrastructure so the app can remain reliable even as providers change model names, pricing, and availability.
Added a new structured output window for code, commands, JSON, SQL, checklists, plans, schemas, and tables.
Completed a public launch security pass covering auth, CORS, HTML sanitization, media downloads, update delivery, and provider route access.
Also In v3.2.5 - Research Surface
Research answers now feel like a first-class part of Virgil instead of a plain chat response. Writers can ask deeper questions and see the sources, progress, and context behind the answer.
Added live research progress so users can see what Virgil is doing while it gathers information.
Added source cards, source rails, and expandable source drawers for richer cited answers.
Created a shared research experience for normal chat and Agent Mode.
Improved the chat renderer so research, creative writing, edit diffs, and structured outputs each get their own dedicated presentation.
Polished the visual system across chat, creative windows, diff windows, screenplay views, and the main app shell.
v3.2.4 - Claude Agent Reliability
Improved Claude-powered Agent Mode so longer writing and editing runs feel smoother and more dependable.
Made agent activity easier to follow across chat, screenplay, and outline windows.
Strengthened the live bridge that keeps multi-window agent sessions in sync.
v3.2.3 - Build And Release Maintenance
Updated Mac and Windows build workflows.
Improved release consistency across desktop platforms.
v3.2.2 - Claude Runtime Fixes
Fixed Claude runtime behavior after the larger Agent Mode updates.
Improved screenplay change-set handling and backend reliability.
v3.2.1 - Storyboard, Video, And Backend Patch
Improved storyboard and video generation flows.
Smoothed generated media handling and UI state.
Added release validation fixes for packaged desktop builds.
v3.2.0 - Style Library, Retrieval, And Video Capabilities
Virgil became more context-aware. This release added the foundation for style memory, richer project retrieval, and broader media generation.
Added style library functionality so Virgil can remember and reuse writing style context.
Added retrieval infrastructure for conversations, documents, outlines, embeddings, and script context.
Expanded video provider capabilities, including xAI/Grok video support.
Improved storyboard, media, and video generation routes.
Refreshed app icons, update progress screens, packaging, and desktop update behavior.
v3.1.8 - v3.1.12 - Backend Release Train
Shipped a fast sequence of backend and desktop patches leading into v3.2.0.
Improved provider route cleanup, versioning, release metadata, and media behavior.
v3.1.7 - Release Polish
Continued desktop release polish after updater and icon improvements.
Kept Mac and Windows packaging aligned.
v3.1.6 - Logo And Updater Fixes
Updated Virgil's branding assets.
Improved packaged desktop updater behavior.
v3.1.5 - Emergency Agent Fixes
Fixed urgent Agent Mode scoping issues.
Corrected Claude configuration.
Stabilized agent behavior after the tool-execution updates.
v3.1.4 - Desktop Updater Reliability
Improved auto-updater reliability across desktop platforms.
Tightened packaged release behavior for Mac and Windows.
v3.1.2 - v3.1.3 - Release Maintenance
Shipped release metadata and package updates following v3.1.1.
Kept the patch line moving while hotfixes were validated.
v3.1.1 - Web Foundation, Cloud Saving, UI Overhaul, And Tool Cancellation
This release pushed Virgil beyond a desktop-only app and toward a more flexible creative workspace.
Added a web-capable foundation so Virgil can run beyond the Electron desktop path.
Added cloud-oriented project and conversation persistence work.
Overhauled major parts of the UI, including chat, project management, video generation, screenplay tools, settings, and app styling.
Improved video editing and video generation across multiple providers.
Added style library functionality for reusable creative voice and tone.
Improved straight-chat tool execution and cancellation across providers.
Fixed AWS, video generation, outline window, backend latency, vector infrastructure, and other release-blocking issues.
v3.1.0 - Multi-Provider Agents
Virgil became a true multi-model writing environment, with stronger Agent Mode and a much broader provider lineup.
Expanded support across Claude, OpenAI, Gemini, Grok, DeepSeek, Qwen, Kimi, and OSS models.
Added model icons and a more polished model selection experience.
Improved autonomous agent execution, writing tools, edit tools, outline tools, communication tools, and web-search tools.
Added richer real-time status indicators so users can see what the agent is doing.
Improved detached windows and live sync across chat, outline, and agent status.
Continued UI polish across chat, video generation, settings, and screenplay editing.
v3.0.1 - Windows Release Fixes
Fixed Windows packaging and startup issues after the v3.0 launch.
Added small UI polish and updated model pricing data.
Improved screenplay multi-edit controls, storyboard styling, and billing metadata.
v3.0 - Video Generation
Virgil moved from screenwriting assistant into visual development partner with the first major video generation release.
Added a dedicated video generation window.
Added backend infrastructure for video routes, media handling, provider pricing, and provider abstractions.
Added provider support paths for OpenAI, Google, Runway, Luma, Kling, Hailuo, and other video systems.
Connected video generation into storyboard and media workflows.
Updated billing and cost tracking for video and model usage.
Improved Gemini UI behavior, chat styling, and app layout.
