BlogNewsUse casesDocsReleases

Changelog

Version History

Follow the evolution of genvi as we build the world's most powerful collaborative, node-based multimodal AI workspace.

v0.7

•Genvi Agent and the Reference Library
majorSeptember 7, 2026

Two ways to stop repeating yourself. Genvi Agent works your canvas directly while you watch, and the Reference Library lets you describe a character, a product or a look once and then call it by name in any prompt, on any canvas.

Genvi Agent

  • Ask for the thing, not the steps. "Add a title card and a call to action at the end" is a complete instruction.
  • It works where you are: Node Studio, both composers, the whiteboard and 3D. Every edit lands in your undo history, because it uses the same controls you do.
  • Anything that spends credits or deletes work stops and asks first. Leave it on Ask, or let it run.
  • Reach it from any canvas with Ctrl+J.

Reference Library

  • Save a character, a place, an object, a style or a sound direction once: a name, a description written for a model, and the pictures that pin it down.
  • Type @nova in any prompt box and it becomes that description and those pictures. Every studio reads a mention the same way, and your wording reaches the model exactly as you wrote it.
  • Bundle a cast or a product range into a set, and one mention brings all of them.
  • Save any generation straight back to the library, then mention it from then on.

Pick your shot before you pay for it

  • Ask Chat Studio for a video and it can draft four opening frames first. Pick the one you want and it animates that.
  • Four frames cost single-digit credits where a video runs into the hundreds, so a composition you did not want is cheap to find out about.
  • None of them right? Generate the video anyway, in one click.

New models, and fair warning before the old ones go

  • Added: GPT-6 Astra, Gemini 3.8 Flash, Veo 3.1, Lyria 3.5, Wan 3.0 Prime, MiniMax H3 Max Turbo, Meshy v7 for 3D, and Muse Image, now the cheapest image model in the catalog.
  • Models on the way out say so. The picker shows the retirement date and what to switch to, and the node repeats it under the model name.
  • When one does shut down, your saved projects keep rendering: we route them to the replacement. The notice exists so the change is never a surprise, because the output does come from a different model.
  • Sora has been withdrawn. OpenAI is closing its video API with no successor. Veo 3.1 and Kling v3 Pro are the closest equivalents.

Generated pipelines you can read

  • Ask Genvi Agent for a pipeline, use the pipeline generator, or drop in a template, and the nodes now arrive properly laid out: real columns, room for the wires, nothing overlapping.
  • The layout settles once the cards have measured themselves, so it fits the nodes you actually got rather than an estimate of them.

Fixes worth naming

  • Large files upload again. Anything over about 4.5 MB was being refused before it reached us, which is why a big video failed with no reason given.
  • Written copy, lyrics and scripts generate properly again on the latest text models.
  • Reasoning Level reaches the model again. On every GPT-5 model it was being sent under the wrong name and quietly ignored, so the control did nothing.
  • Gemini 3 image resolution is honoured again, so the tier you choose is the tier you get and the tier you are billed for.
  • Lyria writes WAV when you ask for WAV. Every render came back MP3 whatever you picked.
  • Five models were sending every run to an id their provider had retired or never published, so those runs failed outright. They now route to live models.
  • Video length, resolution and background save when you change them, instead of reverting a few seconds later.

How we measure the product

  • We now measure how genvi is used, so we can fix what is broken and build what is used. It is first-party only, stored in our own database, and never shared with an analytics vendor or an ad network.
  • It sets no tracking cookie, and signed out nothing is written to your browser at all. The cookie and privacy pages say all of it in full.

v0.6

•The Chat Update
majorAugust 19, 2026

Just say what you want. A new conversational studio turns a sentence into finished images, video, music, voice-over or 3D, and picks the right model for the job so you never have to.

Chat Studio

  • Say it in plain words and get it back. Images, video, music, voice-over, copy and 3D, all from one conversation.
  • No canvas to learn and no models to weigh up. Leave it on Auto and the right one is chosen for you.
  • Want to take the wheel? Every model, and every setting it offers, is one dropdown away.
  • Not quite right? Just reply. It remembers what you were making, so "make it warmer" is a complete instruction.
  • Chase a second idea without losing the first. Branch any message into a chat of its own.
  • Keep a whole campaign in one project, where your characters, colours and formats are understood from the start.

Work from your own images

  • Drop in a photo to steer the result. Upload it, pull it from your files, drag it in, or simply paste.
  • Say how it should be used: the frame a video opens on, a face to keep, or just a look to borrow.
  • See exactly which of your images reached the model, before you spend a credit and after.
  • Curious what it actually asked for? The finished prompt sits right beside the words you typed.

Fits the way you already work

  • Everything you make saves itself to your files, ready to drop straight into the composers.
  • Start something and carry on. Generations keep running while you move between chats.
  • A calmer, more consistent look across every studio, and controls that work properly on touch.

v0.5

•The Composer Update
majorJuly 16, 2026

Animate anything on a real timeline, stack videos picture-in-picture, and design with pro-grade text and effects. Now you can do all of it from your phone.

Video Composer

  • Animate images, text, video and shapes with keyframes, so you can build motion without leaving the timeline.
  • Stack multiple videos at once with picture-in-picture, and mask any layer.
  • Ramp your footage with speed curves for slow-motion and time-stretch.
  • Apply filters and effects to any track and see them carried straight through to your export.
  • Smoother scrubbing and playback, and far more dependable exports.

Image Composer

  • Lay out copy with text boxes, ready-made text styles and text on a curve.
  • Finish images with artistic effects like edge inking, painterly and halftone.
  • Mask layers to the shape beneath them and fill with multi-stop gradients.
  • Draw with pen, pencil, marker and highlighter brushes that feel real, plus a true eraser.

Whiteboard

  • Drop in GIFs without leaving the board, with search powered by GIPHY.
  • Every tool now opens its settings in one panel, so nothing floats over your canvas.
  • New objects land in view, ready to edit.

Editing On Your Phone

  • The image composer now gives you the same tools on mobile that you get on desktop.
  • A video timeline and properties built for touch.

A Cleaner Interface

  • One consistent look across every editor, with the same controls in the same places.
  • Document settings, zoom and save gathered into a single top bar.
  • A redesigned colour picker and sliders throughout the app.

Models & Plans

  • Generate several images in a single run.
  • New text-to-speech voices you can preview before you commit.
  • Subscription plans and credit top-ups.
  • Your free monthly credits now arrive on your signup date instead of the 1st of the month.

v0.4

•The 3D Studio Update
majorJune 20, 2026

Direct 3D mesh integration, HDRI environment mapping, dynamic orientation heuristics, texture AI projection painting, and real-time team workspaces collaboration.

3D Studio Engine

  • Direct integration of 3D studio mode inside node preview elements for fluid viewport editing.
  • Support for GLB and GLTF file uploads directly to the repository and canvas.
  • Built-in HDRI environment lighting with 7 default 2K maps (Sunset, Studio, Night, Outdoor, etc.) for realistic rendering previews.
  • Implemented vertex-based boundary heuristic scoring to automatically calculate front-facing camera coordinates for horizontal and vertical models.
  • Added auto-framing 3/4 angled viewport camera configurations for project card thumbnail generation.

AI Texture Painting

  • Implemented AI-assisted projection texturing directly onto loaded 3D meshes.
  • Added stencil/brush reset view shortcuts to facilitate seam-free coordinate mapping.
  • Resolved UV seam projection bugs that previously jumbled textures on complex mesh maps.

Collaborative Workspaces & Teams

  • Implemented team membership authorization logic inside backend database sync wrappers (cloudGetProject, cloudUpsertProject, cloudRenameProject, cloudDeleteProject).
  • Enabled multi-user read/write/delete sharing so collaborators can edit projects together inside a single workspace, while preserving original owner metadata.

Admin Dashboard & UI Polish

  • Added Credit Config management panel with per-second model run rate configuration.
  • Migrated account and pricing settings into overlay modals with glassmorphic backdrop blurs.
  • Replaced all default browser confirm/alert dialogue boxes with custom, themed React modals and global toast notifications.

v0.3

•Templates & Sharing
majorMay 20, 2026

Introduction of Project Templates, a publisher permission ecosystem, dashboard skeleton load states, and cache-prefetched tab switching.

Project Templates System

  • Migrated template system to published_projects schema to support template versioning and data persistence.
  • Added thumbnail file uploads and custom public file pickers in the project publishing modal.
  • Added template validation checks for workspace import parameters.

Publisher Access Controls

  • Introduced dedicated Library Tab and Publisher management interfaces to control project templates visibility and permission levels.
  • Routed cloud filesystem uploads through publisher assets API.

Dashboard Performance

  • Implemented dynamic interactive canvas skeleton card loading transitions to replace generic spinners.
  • Prefetched template assets on mount for zero-latency, cache-served tab navigation.
  • Added URL-based tab state persistence to preserve user location upon reload.

v0.2

•Whiteboard Canvas & Tools
majorApril 10, 2026

Transitioned whiteboard into a robust infinite canvas with drawing overlays, shape tools, and generation indicators.

Infinite Canvas Tooling

  • Implemented infinite canvas selection boundaries with outline handlers.
  • Added whiteboard drawing, marker, and highlighter layers overlay rendering.
  • Added custom vector shapes, emoji lookups, and rich media embedding directly onto nodes.

UX Improvements

  • Fixed note text-editing cursor jumping by disabling active database sync hooks during typing.
  • Added synthetic heartbeat progress bars for slow image and video generation queues.

v0.1

•First Beta Release
majorMarch 1, 2026

Launch of genvi node-based multimodal prompt-chaining platform, video editor, and multi-layered raster editor.

Node Studio

  • Implemented node-based visual workflow system to connect text inputs, image generator, and video compiler models.
  • Added BYOK (Bring Your Own Key) support so you can use your own AI provider credentials, encrypted with authenticated AES-256-GCM.

Video & Image Composers

  • Added Video Composer timeline with multitrack supports, text layers, audio tracks, and resolution properties.
  • Added Image Composer with layer nesting, adjustment parameters, opacity options, and custom raster brush controls.