What's new in ModelRoom — the features we've shipped, the things we've polished, and the fixes we've rolled out.
Drop a PDF, Word, text, or Markdown brief into the chat and your agent works straight from it — no retyping your docs.
Viewing a room someone shared with you? Copy it into your account and build on it — the original stays untouched.
Running low mid-project? Buy a one-time credit pack without changing your plan — and redeem a referral code for bonus credits.
Seedance 2.5 joins your video models — reference-driven clips from 4 to 30 seconds, twice the length you could generate before.
MiniMax H3 joins your video models — sharper clips with native stereo audio, steered by image, video, and audio references.
Panning, dragging, and selecting stay smooth in rooms packed with nodes, so the canvas keeps up as your work grows.
Flip through everything a node has generated and restore any earlier result — iterate freely without losing the version you liked.
Stop a turn any time and keep what your agent already built — and attach up to ten images per message, phone photos included.
Pick trackpad or mouse controls for panning and zooming, and the canvas now stays put while your agent reworks the room.
Send someone your room link and they can explore the whole canvas in view-only mode — nothing they do changes your work.
Link ModelRoom to external AI clients with a secure one-click sign-in, then create and compose in your rooms without leaving the tools you already use.
Browser notifications let you know the moment a generation finishes or your room agent replies, so you can step away and come back right when it matters.
Pick the resolution for your video generations, with the credit cost updating to match exactly what you choose — so you control the trade-off between quality and price.
Add a brief to a project and your room agent follows it across every room inside — so your goals and guidelines carry through without repeating yourself.
A new media panel gathers every image and video you’ve made — filter by type, scan by date, and jump back to any result — across Recents, Projects, and your creations library.
Upload an audio clip and connect it to video generation as a reference, so the sound you bring helps shape the clip you get back.
Long sessions hold up, interrupted uploads recover, and generation no longer freezes mid-run — and when a run fails, its credits settle correctly so you are not charged for what you did not get.
Describe a video and your room agent storyboards it shot by shot, builds each one, and stitches the final cut — for any kind of video, not just ads. You can keep chatting while it renders.
Drag images straight into the agent chat to anchor the look you want, and hand video generations reference clips to match — your style carries through every output.
Signing in now drops you into your latest rooms, ready to pick up where you left off — with a prompt bar right there for your next idea.
Group rooms into projects, each with its own view and breadcrumb, and create new rooms straight into one. Pin the rooms you keep coming back to, rename them inline, and rename or delete projects anytime — deleting a project moves its rooms back to Main rather than removing them.
A cleaner look across ModelRoom: menus, dialogs, and the command palette now share a frosted-glass style, with a redesigned navigation sidebar and crisper hover states throughout.
Media previews now match the true aspect ratio of your image or video — nothing looks stretched, letterboxed, or cropped — and undo/redo were hardened so reversing a change always lands where you expect.
Spin up a variation just by swapping its inputs — no extra prompt required. A faster path from one concept to many.
Your room agent can now plan multi-step changes from a single chat — adding, editing, or removing parts of the room for you, while quick-reply chips and inline forms keep the back-and-forth fast. No manual wiring required.
New rooms and profiles come with unique, seeded generative artwork — distinct avatars and covers, right out of the box.
Localize a clip in one step: each speaker is re-voiced in their own voice and their lips are re-rendered to match the new audio. One video in, a dubbed video out.
Type an idea right from the homepage and it carries straight into a new room, where the agent picks it up and gets to work.
A creations panel gathers everything you’ve generated in a room, with simple tabs to switch between views and jump back to past results.
Create a reusable voice from a short sample and use it anywhere a voice is accepted — voice-overs, merges, and more. Available on Creator and Enterprise plans.
The agent info bar shows a run’s credit cost up front — and when a video model can add sound, you choose whether to include it, with the price updating to match. No surprises before you commit.
Combine multiple clips on a drag-to-reorder timeline, and even swap the voice on the merged audio — shape both the sequence and the sound before you render.
First time in a room? An interactive walkthrough highlights the essentials step by step, so you reach your first generation without guesswork.
Talk to the variation agent and watch its responses stream in token by token, with the full conversation preserved across sessions.
Generated outputs group into named batches with per-run briefs, and you can download an entire batch in one click.
Create from ready-made ad templates, upload several files at once, and spin up variations straight from a template.