Make your AI videos yours: Meet VideoGen’s new editing features ‘Modify’, ‘Change Subject’, and ‘Reframe’

We need to talk about the regeneration spiral.

You know the one. You prompt a video, it comes back 90% perfect — the composition is right, the movement feels natural, the mood is exactly what you pitched to the client. But the lighting’s a bit flat. Or the subject isn’t quite landing. So you regenerate. And the lighting’s fixed, but now the composition is different. Regenerate again. Better composition, worse movement. Regenerate. You’re now six generations deep and somehow further from where you started than when the first one landed.

VideoGen’s new editing features break that cycle. You can now take a generation you like and actually edit it; work with it — refine, swap, adapt — without torching the parts that already landed. Three new capabilities make this possible: Modify, Change Subject, and Reframe.

What are the new VideoGen Editing features?

VideoGen now includes a set of editing tools that let you refine AI-generated videos after creation. Instead of treating every generation as a finished output you either accept or discard, these tools open up a middle ground — the ability to take what’s working and shape it further.

They give you three new distinct ways to do that. Modify lets you describe changes to an existing video (lighting, style, environment, specific elements) and regenerate with your original as the reference point. Change Subject lets you swap a character or object by uploading a reference image while the rest of the scene holds. And VideoGen Reframe lets you adapt a finished video to different aspect ratios with prompt-guided outpainting, so you can lock in the creative first and handle format requirements after. Each one addresses a different friction point in the AI video workflow, but they share the same principle: your generation has value, and your tools should help you build on it.

What runs through all three is a focus on four things. Control — the ability to direct specific changes rather than hoping a fresh generation gets closer. Consistency — preserving what already works while changing what doesn’t. Collaboration — working with your AI outputs iteratively instead of judging them pass/fail. And craft — the kind of intentional, detail-level refinement that turns a good generation into a usable deliverable.

Key editing features within VideoGen

Each feature targets a different moment in the creative process — the point where you know what needs to change but until now had no way to change it without starting over. Here’s how they work and where they fit into your workflow.

1. Modify: the conversation your AI videos have been missing

Modify turns your existing video into the starting point for the next one. When you describe what you want changed — shift the lighting to golden hour, restyle the environment, adjust a specific element in the frame — the AI is working from your video as a direct input reference, not interpreting your text prompt in isolation and hoping it lands somewhere similar.

The practical difference is immediate. Composition carries through. Movement stays consistent. You’re sculpting something that exists rather than describing something you hope will exist again. Prompt a cityscape to shift from noon to dusk, and the buildings, the angles, the depth all hold while the light transforms around them. Restyle a realistic street scene into something illustrated, and the spatial relationships survive the translation.

2. Change Subject: everything stays, except the star

You’ve generated a scene that works beautifully. The environment, the camera movement, the pacing, the lighting — all of it. There’s just one problem: the person in it needs to be someone else. Or the product needs to change. Or you’re building a campaign that requires the same scene with five different characters.

Previously, this meant five separate generations and the quiet hope that the AI would reconstruct the same scene five times. (Narrator: It would not.)

Change Subject lets you upload a reference image of a character or object and swap it into your existing video while everything else holds steady. The environment stays. The lighting stays. The camera movement stays. You’re replacing the lead actor, and the set doesn’t notice.

This matters most for creatives producing content at scale: campaigns with multiple brand personas, product variations shot in consistent environments, or character-driven series where visual continuity across scenes is the difference between professional and patchwork. Upload your reference (up to 10MB), prompt the changes, and the swap preserves what you built around it.

3. Reframe: because creative decisions shouldn’t be hostage to aspect ratios

Here’s a workflow that will feel painfully familiar to anyone producing video for more than one platform: you generate something you love in 16:9, and then you remember you also need 9:16 for TikTok, 1:1 for Instagram, and 4:3 for the pitch deck you forgot about. So you regenerate three more times, each one a fresh gamble on whether you can recapture what made the original work.

Reframe asks a reasonable question: why are you choosing your aspect ratio before you know whether the video is any good?

Instead, you generate your video, land on something you’re genuinely happy with, and then adapt it to any of seven aspect ratios — 16:9, 9:16, 1:1, 4:3, 3:4, 21:9, and 9:21. Audio is preserved, and output quality stays production-ready. 

And here’s where things get really cool: when a new aspect ratio needs additional frame space (going from 9:16 to 16:9, for example), Reframe uses prompt-guided outpainting. You describe what should fill that expanded area rather than crossing your fingers. The difference between “I told the AI to extend my beach scene with more shoreline and sky” and “the AI filled 40% of my frame with whatever it felt like” is the difference between a usable deliverable and another trip back to the generation screen.

Reframe also works on your generation history, turning every video you’ve ever created in VideoGen into multi-format source material. That back catalogue of 16:9 generations? Now it’s also a library of vertical, square, and cinematic widescreen content. No regeneration required.

Most AI video platforms let you pick your aspect ratio upfront, and that’s where the conversation ends. Reframe lets you pick it after — once you actually have something worth adapting. 

What this means for your workflow

These three features share a common thread: they all assume your generation has value worth preserving. That sounds obvious, but it runs counter to how most AI video tools have worked so far, where every output is treated as disposable — a draft you’ll probably throw away on the way to the next draft.

VideoGen’s new editing tools treat your generations as creative assets you can build on. Modify lets you direct how they evolve. Change Subject lets you vary them with precision. Reframe lets you deploy them everywhere your content needs to live. The through-line is control; the kind that turns AI video from a creativity slot machine into something you can genuinely craft with.

That shift, from generating to iterating, is where AI video gets genuinely useful for professional creative work. And we’re just getting started.

Try VideoGen‘s new editing features