Remake video,
without reshooting.

Aufheben AI develops generative AI for high-fidelity local editing of existing videos. Replace backgrounds, remove unwanted elements, and substitute signs, screens, and objects — all at original capture quality, without a reshoot.

See examples

Solution

Remake

Video Local Editing Engine

No reshoot required.

Edit parts of existing videos at high fidelity without disturbing the rest. Replace backgrounds, remove unwanted objects, swap text and logos, or anonymize faces — all without a reshoot. Our proprietary Inversion technology and Flow Matching preserve temporal consistency while delivering natural edits.

MODEL

From 50 MB

SPEED

10 fps

TARGET

Video

KEEP:Temporal consistencyOriginal textureLighting / shadows
EXTRACT:Edit regionMask
PLANNED:3D surface editingHigh-resolution supportLive video supportAudio sync
Background replacement
Object removal (clutter cleanup)
Text & logo substitution
Face anonymization (Face)
3D surface editing
Real-time live video editing
Request a demo →

Vivify

Image/Video AI Model Development

Custom models for your domain.

Contract development of domain-specific CV models for image/video data. Visual inspection, image classification, object detection, segmentation — OSS base + fine-tuning + model compression, delivered to run in your environment in as little as 2 weeks. Led by a doctoral CV engineer from the University of Tokyo.

Image classification / Object detection
Visual inspection (defect/anomaly)
Model compression / Edge deployment
Segmentation

Compose

Coming soon

Video Composition Agent

Assemble videos from raw assets.

An agent that automates material composition and editing for production workflows such as recruiting short videos. Combine multiple video/image assets, generate captions, perform cut editing and color correction — all from natural language instructions. Coming soon as a monthly subscription.

Natural language composition
Multi-asset automatic composition (video/image)
Caption / subtitle generation
Cut editing & color correction
Recruiting short video templates
Join early access →

Demo

Before / After: Video Local Editing

Edit parts of existing footage at high fidelity without disturbing the rest. Background replacement, object removal, and text substitution — all from a unified pipeline.

Legacy
Original
Remake

Recognition retention (Original = 100%)

Edit region fidelity

Legacy
5.7%
Remake
99.9%

Non-edited region preservation

Legacy
0.9%
Remake
83.3%
Request a demo →

Send us a sample video and we'll return an edited sample.

Use Cases

Use Cases

Background replacement

Challenge: Office layout changed after filming, and the video no longer matches current environment

Replace only the background while preserving actors and motion. Swap in the latest office setting — no reshoot, fraction of the cost.

Remake

Removing confidential content

Challenge: Whiteboards, monitors, or documents in the background show confidential information and block public release

Remove or replace only the sensitive elements naturally, preserving the rest. Ready for external use and recruiting campaigns.

Remake

Face anonymization of cast

Challenge: Former employees or unauthorized individuals in the video block continued use

Replace faces with natural alternatives while preserving video structure. Extend the lifespan of existing video assets.

Remake

Plans

Plans

The optimal deployment method for your challenges and phase.

Video Editing (for production companies)

For production companies producing recruiting, explainer, or advertising videos. Background replacement, object removal, and text substitution on pre-delivery footage. Integrates into existing production workflows.

From ¥100K / video
Inquire

Direct Projects / One-off Editing

Direct work with companies providing their own videos. Partial editing for recruiting video revisions, explainer video updates, and archive footage reuse.

From ¥200K / video
Request

Remake API

Coming soon

Send video via REST API and receive edited results. Easy integration into existing video production pipelines. Pay-per-use, ideal for continuous usage.

Coming soon
Pre-register

Compose Agent

Coming soon

Video material composition agent for recruiting short video production. Natural-language-driven asset selection, composition, caption generation, cut editing, and color correction. Coming as a monthly subscription for video production companies and recruiting agencies.

Subscription / Coming soon
Join early access

Vivify Custom Model

Contract development of domain-specific video/image AI models. Visual inspection, object detection, segmentation — OSS base + fine-tuning + compression. Deployable in your environment.

From ¥300K / 2 weeks
Inquire

FAQ

FAQ

Technical

What types of videos can you edit?

+

Coverage depends on video resolution, length, and editing target. Currently optimized for short videos (~2 min) with background replacement, object removal, and text substitution. Complex 3D surface editing and longer videos are on our roadmap. Send us a sample and we'll evaluate feasibility.

Technical

How is this different from AI video generation?

+

Remake edits parts of existing videos. Unlike Runway, Veo, and other generation models that create new footage from scratch, Remake starts from your existing video and edits only the necessary regions at high fidelity. The valuable parts — actors, motion, context — are fully preserved.

Technical

How do you handle temporal consistency (flicker)?

+

Our proprietary Inversion technology and Flow Matching preserve frame-to-frame consistency. There are limits for fast motion and complex lighting conditions, and we evaluate feasibility upfront for such cases.

Ethics

How is this different from deepfakes?

+

Remake is an editing service for videos where the rights holder, videographer, and subjects have granted consent. It is not used to generate false videos impersonating others or to modify videos of individuals without permission. We only serve use cases aligned with our ethical guidelines.

Delivery

What's the turnaround time?

+

For one-off editing projects, 2-7 business days per video is typical. Continuous projects and specialized editing are quoted individually.

Delivery

Can I share confidential videos?

+

Yes, under NDA. Videos are processed in secure environments and promptly deleted after delivery. On-premises processing is also available on request.

Ethics

Responsible Video Editing

As a provider of video editing AI technology, we uphold ethical development and use of the technology.

Proper Rights Verification

We operate on the premise that videos provided to us have the necessary rights (copyright, portrait rights, etc.) to be edited.

Use Case Restrictions

Not available for generating false videos impersonating others or modifying footage of individuals without their consent.

Data Minimization

Only minimum data handled. Promptly deleted after processing. NDA and on-premises processing available.

Transparency on Limits

We proactively disclose the types of videos we can handle and the limits of our editing capability, and evaluate feasibility upfront.

About

About

Mission

Remake video, without reshooting. We build generative AI that edits parts of existing videos at high fidelity — preserving what matters, changing only what needs to change.

Team

Our team combines doctoral-level generative model research from the University of Tokyo with engineering and business design from Kyoto University. We specialize in the intersection of computer vision, privacy-preserving AI, and production deployment.

Video has become the communication asset of choice across every corporate function. Yet revising and diversifying videos after filming still requires expensive reshoots or large-scale re-edits.

We develop generative AI that edits parts of existing videos at high fidelity. Replace backgrounds, remove unwanted content, swap signs and screens — all at original capture quality.

The core is our CTO's deep generative model research at the University of Tokyo. We extended the image Inversion technology accepted at WACV 2025 to video, using Flow Matching to preserve temporal consistency while naturally remaking only parts of the footage.

Building on the video AI foundation we developed for Remake, we are now also developing Compose, a video composition agent. Our goal is to automate, through natural language instructions, the material composition and editing work that occurs daily in production workflows such as recruiting short videos.

Our goal: fundamentally lower corporate video production costs, and maximize the lifespan of every video shot.

代表取締役社長 米山瑛人

Company Info

CompanyAufheben AI
FoundedApril 2026
LocationShibuya, Tokyo
BusinessGenerative AI for video editing (composition agent & local editing); image/video AI model development

For Investors & Partners

Aufheben AI develops generative AI specialized in Real-video Local Editing. Our core technology combines diffusion model Inversion (WACV 2025) from our CTO's deep generative model research at the University of Tokyo (Matsuo-Iwasawa Lab) with Flow Matching for temporal consistency control. Unlike Runway, Pika, and other generation-focused video AI, we differentiate in a specific technical niche: preserving original-footage fidelity while editing only the necessary regions. Initial targets are recruiting/explainer video production companies, advertising agencies, and media companies, with a phased expansion to API and SaaS offerings.

Request Materials →info@aufheben.ai

Contact

Contact

Video editing requests, demo inquiries, plan consultations, materials — feel free to reach out.