3DFY Prompt: AI-powerd 3D models generation at scale. Use 3DFY Prompt in your daily work to empower your 3D design capabilites, or expose text-to 3D to your clients via our API access.3DFY.ai was established with a clear mission: to develop the first technology that transforms text prompts into realistic, high quality 3D models at scale. We are giving 3D creators a tool that they can use in daily work to empower their creativity, and lift some of the burden of daily grind.⚡Key featuresGenerate unique models within a few minutes.Currently we support a few sample object categories. We constantly add more and improve existing ones!PBR textures (diffuse, metallic, roughness, normal, alpha, emission maps available now, soon expandend with AO maps)Adjustable mesh quality/vertex count.Models are divided into semantically meaningful parts, ready for download (.glb, .blend, .fbx) and further editing or import into Unity.⚡Currently supported categoriesSwords, Shields, Axes, Sofas, Tables, Table Lamps, Ottomans, Cutlery (spoons, forks, knives)New categories coming soon👉Ethical AI:We at 3DFY.ai care about ethical AI. We don’t use data from the internet that may cause legal or ethical problems. We create our own data in-house, so you can trust our technology.We're excited to offer our customers a new and innovative way to bring their ideas to life, without copyright or ethical concerns.🦾3DFY Prompt APPYou can generate 3D Models directly from our website after logging in.A newly registered user is entitled to ten (10) 3D model generations credits free of charge. To continue generating and/or downloading models you will need to purchase a relevant packageOur app is designed to be user-friendly and accessible, even for those without 3D modeling experience.🦾3DFY Prompt API (initial preview)The API provides the same model generation capabilities as the 3DFY Prompt web application. Experiencing the web application will give you a good idea of what to expect from the API.It is early access at this point. This primarily means that we do not have a self-service onboarding page. The current process is that you purchase API credits and a team member will issue an API key for you and update API credits in your account.You need to have a 3DFY Prompt account in order to purchase API credits.API credits are offered at a special flat rate of $1 per model generation.To gain access to API preview look for prompt-api channel on our discord server.Unity plugin for integration is currently under development.👉Item categories3DFY.ai’s text-to-3D technology offers a scalable AI-based solution which generates 3D models comparable to those produced by human 3D modelers. The tradeoff for the model quality comes in the form of a category-bound solution, meaning only objects within a range of defined classes can be generated. However, with a continuously expanding range of available categories, this limitation will quickly diminish with time.👉Affordable 3D generationWe use unique generative AI technology to significantly reduce the human labor required to create 3D models and save costs, making 3D modeling accessible to everyone regardless of their budget.While accuracy may vary depending on the complexity of the text prompt, we're committed to ensuring that our models meet contemporary industry quality standards.❓How it worksOur approach to 3D generation is different from other methods that optimize a Neural Radiance Field (NeRF) as an intermediate 3D representation for a given text prompt.Instead, we use a trained encoder-decoder network that extracts a compact and interpretable representation of the input prompt, which we call Object Code. This Object Code is then used by an encoder component along with a set of control parameters to synthesize the 3D asset using a combination of machine learning and computer graphics operations. Our models are trained per object category, which gives them a strong object category prior and ensures they always generate a plausible object within the category.❓How we source our dataTo train our neural networks, we use large datasets of 3D models rather than captioned images sourced from the public internet, as it is common practice for diffusion models. However, sourcing these 3D models can be a difficult and costly task, with issues related to variable quality and multiple formats. To overcome these challenges, we generate our own training data using a combination of machine learning and computer graphics technologies. By doing so, we are able to control the sample distribution of our datasets and ensure that our trained network performs at a satisfactory level. Additionally, since all of our training data is self-generated, we avoid issues related to data quality and bias, and can guarantee that our data is ethically sound. We are proud of our dataset generation capabilities and offer it as a service to our customers.❓How to reach us💬 find us on our Discord server: https://discord.gg/CKu58ZrRQx,✉️ for support you can reach us at: contact@3dfy.aiOur solution comes with its own custom EULA. You can find the details in our terms and conditions.👉 What it is and what it is not3DFY Prompt is a unique technology that allows you to generate 3D models from text prompts. While it may be similar to other text-to-image and text-to-3D generators, it has several key differences. For example, it is category-based, meaning you first select an object category and then provide a prompt to receive a valid object within that category. Our technology is geared towards functionality rather than creativity, so we encourage you to use functional and succinct descriptions.Additionally, we generate 3D models at a high-quality standard, which includes dividing the assets into structural parts, having excellent meshing topology, and using Physically Based Rendering (PBR) material textures. You can even select the file format, LOD, and texture resolution when you download a model from 3DFY Prompt.Overall, 3DFY Prompt is not just another diffusion-based text to 3D generator, but a unique and powerful tool for creating realistic and functional 3D models.👉 CapabilitiesGenerate unique models within a few minutes.Currently we support a few sample object categories. We constantly add more and improve existing ones!PBR textures (diffuse, metallic, roughness normal, alpha, emission maps available now, soon expandend with AO maps).Adjustable mesh quality/vertex count.Models are divided into semantically meaningful parts, ready for download (.glb, .blend, .fbx) and further editing or import into Unity.👉 Best practicesAs the developers of the 3DFY Prompt technology, we recommend using clear and concise prompts to achieve the best results. Our system is designed to create realistic and functional 3D models, so using long and complex sentences or adding implausible features may hinder the system's ability to generate the desired outcome. Instead, it's best to describe what you want in simple, day-to-day words.For example, describing a gentleman's Chesterfield sofa with two seats and a wraparound back with short wooden feet will produce a stunning result. While we encourage creativity, it's important to keep in mind the limitations of our technology. We generate 3D models based on real-world manifestations of the selected category, so using analogies and creative prompts that still fit within the realm of possibility will produce the best results.We understand that generating a 3D model can take several minutes, which is why we keep all of your creations under "My Models" for you to browse, review, and download at any time. You can even open 3DFY Prompt in multiple browser windows or tabs to generate additional items while waiting for your models to generate.Ethical AIWe at 3DFY.ai care about ethical AI. We don’t use data from the internet that may cause legal or ethical problems. We create our own data in-house, so you can trust our technology.We're excited to offer our customers a new and innovative way to bring their ideas to life, without copyright or ethical concerns.Affordable 3D generationWe use unique generative AI technology to significantly reduce the human labor required to create 3D models and save costs, making 3D modeling accessible to everyone regardless of their budget.While accuracy may vary depending on the complexity of the text prompt, we're committed to ensuring that our models meet contemporary industry quality standards.How it worksOur approach to 3D generation is different from other methods that optimize a Neural Radiance Field (NeRF) as an intermediate 3D representation for a given text prompt.Instead, we use a trained encoder-decoder network that extracts a compact and interpretable representation of the input prompt, which we call object code. This object code is then used by an encoder component along with a set of control parameters to synthesize the 3D asset using a combination of machine learning and computer graphics operations. Our models are trained per object category, which gives them a strong object category prior and ensures they always generate a plausible object within the category.How we source our dataTo train our neural networks, we use large datasets of 3D models rather than captioned images sourced from the public internet, as inis common practice for diffusion models. However, sourcing these 3D models can be a difficult and costly task, with issues related to variable quality and multiple formats.To overcome these challenges, we generate our own training data using a combination of machine learning and computer graphics technologies. By doing so, we are able to control the sample distribution of our datasets and ensure that our trained network performs at a satisfactory level.Additionally, since all of our training data is self-generated, we avoid issues related to data quality and bias, and can guarantee that our data is ethically sound. We are proud of our dataset generation capabilities and offer it as a service to our customers.
