Product Update: Image Lab, Avatar Videos, and Automation APIs - May 5, 2026

These past two weeks have been one of the biggest stretches of new features we've shipped at ngram. There's a brand-new Image Lab for standalone image generation, a full Avatar video mode with face and voice pairing, three new automation integrations for developers, and a long list of quality-of-life improvements across the video creation flow. Here's everything that's new.

What's New

Generate Images Without Making a Video

If you've ever wanted to generate a quick product mockup, social graphic, or brand visual without going through the full video workflow, Image Lab is for you. It lives at `/app/image-lab` - a standalone image generation tool with its own templates, model selector, and editing history.

Pick from business templates or write a free-form prompt. Choose from eight AI models including Grok Imagine, Reve 1.5, Flux 2 Pro, Flux Kontext Max, and Seedream v4.5 - each with different strengths for photorealism, illustration, or brand-style generation. There's an "On Brand" style mode that pulls your brand colors, fonts, and logo into the generation. Once you've got an image you like, edit it further with reference images or style tweaks, and your full history is saved in a sidebar so you can revisit past generations.

Free and Basic plan users get a watermarked version alongside the clean original, so everyone can try it out.

Create Presenter-Led Avatar Videos

Avatar is a new animation mode that puts an AI-generated presenter front and center in your video. Instead of stock footage or motion graphics illustrating your script, you get a talking-head presenter delivering your content directly to camera.

Start by picking a face from the new Face Library - a curated set of presenter avatars, or upload your own custom face from the Faces settings page. Then choose a voice. ngram automatically filters voices by gender compatibility, so you won't accidentally pair a male face with a female voice. The before-start dialog walks you through avatar and voice selection together in a two-column layout before you begin.

The avatar mode has its own preview tile with a sample video, so you can see what the output style looks like before committing.

Turn Screen Recordings into Polished Videos

Screencast mode is now available to everyone. Previously limited to internal testing, it's now a first-class animation option in the wizard alongside AI Motion Graphics and Flow.

Upload a screen recording, and ngram's screencast tools handle the rest - dedicated script writing that describes what's happening on screen, scene-by-scene animation that highlights key moments, and editing tools designed specifically for screencast content. It's built for product demos, tutorials, and walkthroughs where you want the actual software interface as the visual, not AI-generated imagery.

Pick Your Background Music Before You Start

Background music is no longer a black box. There's a new music picker in the chat input area where you choose the energy level - Calm, Upbeat, or Energetic - and preview sample tracks before your video starts generating.

Calm gives you ambient, low-key underscore. Upbeat adds tempo and rhythm for explainers and product tours. Energetic pushes it further with driving beats - great for product launches or ads. The AI uses your selection to guide the final music composition, so what you hear in the preview is close to what you'll get in the video.

Control Your Videos with AI Coding Agents

ngram now has a public MCP server - meaning AI coding assistants like Claude, Cursor, Windsurf, and others can create and manage ngram videos programmatically. Point your MCP client at `https://mcp.ngram.com`, connect your API key, and your AI agent can create videos, check project status, and browse the template catalog without you touching the ngram UI.

The server is hosted at mcp.ngram.com and listed on Smithery for easy discovery. If you're building automated content pipelines or integrating video generation into developer workflows, this is the entry point.

Automate Video Creation with Make.com and n8n

Two new no-code integrations join the ngram automation toolkit. The Make.com integration lets you build scenarios that create videos, poll for status, and trigger downstream actions when a video completes or fails - all through Make's visual workflow builder.

The n8n community node does the same for n8n users: create videos, check status, and set up webhook triggers for video-ready and video-failed events. Both integrations use ngram's public API and support the full video creation flow.

Upload PDF Slide Decks Directly

You can now drag a PDF into the slides-to-video wizard and ngram converts each page into a slide image automatically. Previously you needed to export your presentation as PowerPoint first - now PDFs work natively. Each page becomes a slide frame that flows through the same keyframe animation pipeline as PowerPoint uploads.

Save and Manage Presenter Faces

The new Face Library mirrors the voice library pattern. Browse curated presenter faces, or upload your own custom face from the Faces settings page. Your selected face appears as a pill on the input bar (just like the voice pill) and flows into the conversation - the AI knows which presenter to use without you having to describe them each time.

Custom voices are now also recognized in the face+voice pairing flow, so if you've cloned your own voice, it works seamlessly with the face selector.

Mark Up Any Storyboard Frame for Precise Edits

Markup mode (previously "Suggest Edits") lets you click directly on any storyboard frame and leave a visual note - "make this text bigger," "swap the background image," "move the logo to the right." The AI reads your markup and applies targeted changes to that specific frame instead of regenerating the whole scene.

The markup button lives in a header strip above the asset viewer, and there's a comments-list popover for managing multiple markups. For video and motion graphics scenes, you can leave feedback on specific frames. The AI's proceed suggestions appear as a next-step CTA in the header, so the edit-review-approve cycle stays in one place.

Improvements

Redesigned Storyboard Timeline

The storyboard tab got a visual overhaul. Scenes are now grouped into bordered scene-boxes with a "Scene N - duration" header, replacing the flat tile strip. The active scene gets a tinted border, and individual tiles use a subtle inner outline instead of the previous ring treatment. Loading states are cleaner too - a centered progress ring with percentage and a tool-step caption replaces the old shimmer overlay.

Smarter "Before We Start" Setup

The preflight dialog that appears before video creation now shows a "Recommended" badge on the animation mode it thinks best fits your content. Hover over the badge to see why - "Best for product demos with screen recordings" or "Recommended for talking-head explainers." The recommendation comes from a Gemini-backed preflight analysis of your input.

Silent Videos and Caption Control

Two related changes: you can now explicitly choose "No voiceover" when setting up a video, and you can toggle captions on or off at any point during the conversation. Previously, captions were always on by default - now they're off by default, so you add them when you want them rather than removing them when you don't.

Pronunciation Preview

Before committing to a voice, you can now preview how specific names, products, or technical terms will sound. There's a preview button in the before-start modal that plays a short ElevenLabs-generated sample using your selected voice.

Undo and Fork from Any AI Response

Every assistant message now has undo and fork icons. Undo rolls back to the state before that response. Fork creates a branch point - useful when you like where the video was heading two turns ago but want to try a different direction from there.

More Voices and Languages

The voice catalog is significantly expanded with more ElevenLabs voices and multilingual support. There's a language selector in the voice picker, and your chosen language flows through the entire TTS pipeline - narration, voiceover edits, and minor corrections all respect the selected language.

"AI Motion Graphics" Is Now the Default

The animation option previously called "Mix (Beta)" is now "AI Motion Graphics" and it's the default choice when you start a new video. The reordering puts AI Motion Graphics first, Flow second, and Screencast third - reflecting where most users get the best results.

Transition Previews in Chat

Scene transition videos now appear directly in the chat panel as you're building your video. Previously you had to switch to the video tab to see transitions - now they render inline with a label so you can preview them without leaving the conversation.

Smoother Video Playback

The video preview player got a major under-the-hood upgrade. A new v2 player stack uses Remotion's premount and preload capabilities to eliminate two longstanding playback issues: audio that was silent for the first few seconds of playback, and visible lag when transitioning between scenes. Videos now play smoothly from the first frame.

Bug Fixes

• Video filenames fixed - exported videos were saving as "Untitled_project_Untitled_video.mp4" regardless of your project name. They now use the correct project and video title.
• PowerPoint uploads restored - PPT/PPTX files weren't appearing in the asset library after upload. They now show up correctly, and the slideshow creation flow properly splits slides into individual frames.