AI-Powered Customisable Visual Content Creator for Marketers

AI Decisions revolutionizes generative AI by enabling precise, instance-level object manipulations in images while maintaining realism. This innovation addresses key limitations, unlocking new potentials across high-growth industries and transforming image generation from textual prompts with fine-grained control.

Video about technology

Our Vision:

Our project will revolutionize generative AI with advanced instance-level object manipulations in images, addressing current model limitations and unlocking new potential across high-growth industries.

Problem:

Limitations of current generative models.

Despite their advantages, current generative models face significant limitations:

Lack of precise instance-level control

Loss of detail and realism during adjustments

High computational costs for fine-tuning

The target customers:

Companies using generative models are automating creative tasks, increasing efficiency and innovation across all sectors + the developers (we can provide our APIs so that they can develop their own applications on top of it):

Media and Entertainment: Special effects and realistic characters

Marketing and Advertising: Diverse promotional materials

Education: Immersive educational content

Healthcare: Detailed anatomical images and surgical simulations

Solution:

AI Decisions is creating a revolutionary solution – the System that allows precise manipulation of specific Generative AI objects at the instance level without compromising quality:

Instance-level control: Specific object manipulation with high fidelity

Efficiency: Reduced computational costs and time

User-Friendly: Intuitive interfaces for non-technical users

Video of results here. We’ll be uploading a video of how this happens soon.

Video of results

Target market:

Companies using Genitive AI to create  any objects (arts, music etc.).

TAM: £53 BN and 30% CAGR*

SAM: £27 BN

SOM: £1 BN

Competitors:

Midjourney

Current state-of-the-art generative models such as Midjourney are not able to edit already generated image, especially complex attributes such as position.

Adobe Photoshop

Adobe Photoshop can do something similar, but it is not an open-source app and requires a lot of manual actions.

DALL-E

Open AI (DALL-E, CLIP) Strong in text-to-image generation with models like DALL-E, but can be enhanced with finer control using CLIP.

BERT, LaMDA, Gemini

Google (Imagen, Parti) Superior in high-resolution image generation with Imagen,  but integration of advanced segmentation could offer more precise manipulations.

GauGAN, Clara

Nvidia (GauGAN) Excels in generating realistic images from sketches, but lacks direct text-based manipulation capabilities.

ModelsLab

ModelsLab does not enable precise manipulation such object repositioning, resizing, removal and addition.

Different research methods

Self-correcting LLM-controlled Diffusion (SLD) does not enable precise manipulation (only through text).

DragonDiffusion require manually define a segmentation mask or a bounding box.

DiffEdit enables only a limited set of manipulations (through text only), not instance-level.

Traction:

The prototype is ready and available upon request. We are actively working on reducing both the time to generate changes and the required power. An updated system is planned for release in 2025.

If you are interested in investing, please write to us

LEGAL INFORMATION

107B Cottenham Park Road
London
SW20 ODS
United Kingdom

AI DECISIONS – 2024

AI-Powered Customisable Visual Content Creator for Marketers

Our Vision:

Our project will revolutionize generative AI with advanced instance-level object manipulations in images, addressing current model limitations and unlocking new potential across high-growth industries.

Problem:

Limitations of current generative models.

Despite their advantages, current generative models face significant limitations:

Lack of precise instance-level control

Loss of detail and realism during adjustments

High computational costs for fine-tuning

The target customers:

Companies using generative models are automating creative tasks, increasing efficiency and innovation across all sectors + the developers (we can provide our APIs so that they can develop their own applications on top of it):

Media and Entertainment: Special effects and realistic characters

Marketing and Advertising: Diverse promotional materials

Education: Immersive educational content

Healthcare: Detailed anatomical images and surgical simulations

Solution:

AI Decisions is creating a revolutionary solution – the System that allows precise manipulation of specific Generative AI objects at the instance level without compromising quality:

Instance-level control: Specific object manipulation with high fidelity

Efficiency: Reduced computational costs and time

User-Friendly: Intuitive interfaces for non-technical users

Video of results

Target market:

Companies using Genitive AI to create any objects (arts, music etc.).

TAM: £53 BN and 30% CAGR*

SAM: £27 BN

SOM: £1 BN

Competitors:

Midjourney

Current state-of-the-art generative models such as Midjourney are not able to edit already generated image, especially complex attributes such as position.

Adobe Photoshop

Adobe Photoshop can do something similar, but it is not an open-source app and requires a lot of manual actions.

DALL-E

Open AI (DALL-E, CLIP) Strong in text-to-image generation with models like DALL-E, but can be enhanced with finer control using CLIP.

BERT, LaMDA, Gemini

Google (Imagen, Parti) Superior in high-resolution image generation with Imagen, but integration of advanced segmentation could offer more precise manipulations.

GauGAN, Clara

Nvidia (GauGAN) Excels in generating realistic images from sketches, but lacks direct text-based manipulation capabilities.

ModelsLab

ModelsLab does not enable precise manipulation such object repositioning, resizing, removal and addition.

Different research methods

Self-correcting LLM-controlled Diffusion (SLD) does not enable precise manipulation (only through text).

DragonDiffusion require manually define a segmentation mask or a bounding box.

DiffEdit enables only a limited set of manipulations (through text only), not instance-level.

Traction:

The prototype is ready and available upon request. We are actively working on reducing both the time to generate changes and the required power. An updated system is planned for release in 2025.

If you are interested in investing, please write to us

Companies using Genitive AI to create  any objects (arts, music etc.).

Google (Imagen, Parti) Superior in high-resolution image generation with Imagen,  but integration of advanced segmentation could offer more precise manipulations.