Amazon Titan in Amazon Bedrock

High-performing foundation models from Amazon

Benefits

High-performing image, multimodal, and text models to power a broad range of generative AI applications, such as content creation, image generation, and search and recommendation experiences.
Enhance search accuracy and improve personalized recommendations.
All Amazon Titan FMs offer built-in support for the responsible use of AI by detecting and removing harmful content from the data, rejecting inappropriate user inputs, and filtering model outputs. All images generated by the Amazon Titan Image Generator model contain an invisible watermark by default to help mitigate harmful content generation and reduce the spread of misinformation. Watermark detection is an innovative technology, and AWS is among the first major cloud providers to widely release built-in watermarks for AI image outputs.
Fine-tune Amazon Titan models with your own data to customize the model and perform organization-specific tasks. Easily customize the model responses and deliver accurate domain specific responses by connecting FMs to your data sources using Retrieval Augmented Generation (RAG).

Meet Amazon Titan

Exclusive to Amazon Bedrock, the Amazon Titan family of models incorporates Amazon’s 25 years of experience innovating with AI and machine learning across its business. Amazon Titan foundation models (FMs) provide customers with a breadth of high-performing image, multimodal, and text model choices, via a fully managed API. Amazon Titan models are created by AWS and pretrained on large datasets, making them powerful, general-purpose models built to support a variety of use cases, while also supporting the responsible use of AI. Use them as is or privately customize them with your own data.

Use cases

Use Titan Text models to improve productivity and efficiency for an extensive range of text-related tasks, such as creating copy for blog posts and web pages, classifying articles into categories, open-ended Q&A, conversational chat, information extraction, and more.

Use Titan Text models to get concise summaries of long documents such as articles, reports, research papers, technical documentation, and more to quickly and effectively extract important information.

Use Titan Multimodal Embeddings and Titan Text Embeddings to power more accurate and contextually relevant multimodal search, recommendation, and personalization experiences for end users.

Enable content creators with rapid ideation, customization and iteration resulting in high efficiency image generation and image editing. Customers in industries like advertising, e-commerce, and media and entertainment can generate realistic, studio-quality images in large volumes and at low cost, using natural language prompts.

Deliver more up-to-date and accurate results for user queries by connecting FMs to your data sources. Extend the already powerful capabilities of Titan models and make them more knowledgeable about your specific domain and organization.

Model versions

Amazon Titan Text Premier

Amazon Titan Text Premier is a powerful and advanced large language model (LLM) within the Amazon Titan Text family, designed to deliver superior performance across a wide range of enterprise applications. This model is optimized for integration with Amazon Bedrock Agents and Amazon Bedrock Knowledge Bases, making it an ideal option for building interactive generative AI applications that can use your APIs and interact with your data. 

Max tokens: 32K

Languages: English

Fine-tuning supported: Yes (preview)

Supported use cases: Chat, chain of thought reasoning, open-ended text generation, brainstorming, summarization, code generation, table creation, data formatting, paraphrasing, rewriting, extraction, and Q&A.

Read the blog

Watch Titan Text Premier demo

Amazon Titan Text Express

LLM offering a balance of price and performance.

Max tokens: 8K

Languages: English (GA), 100+ languages available (preview)

Fine-tuning supported: Yes

Supported use cases: Open-ended text generation, brainstorming, summarization, code generation, table creation, data formatting, paraphrasing, chain of thought, rewriting, extraction, Q&A, and chat.

Watch Amazon Titan Text demo

Watch re:Invent session

Amazon Titan Text Lite

Cost-effective and highly customizable LLM. Right-sized for specific use cases, ideal for text generation tasks and fine-tuning.

Max tokens: 4K

Languages: English

Fine-tuning supported: yes

Supported use cases: Summarization and copywriting.

Watch Amazon Titan Text demo

Watch re:Invent session

Amazon Titan Text Embeddings

Model that translates text into vector representations (embeddings).

Max tokens: 8K

Languages: 25+ languages

Fine-tuning supported: no

Embeddings: 1,536

Supported use cases: Text retrieval, semantic similarity, and clustering.

Read the blog

Amazon Titan Text Embeddings V2

Embeddings model optimized for high accuracy and retrieval performance at smaller dimensions for reduced storage and latency.

Max tokens: 8K

Languages: 100+ in pretraining

Fine-tuning supported: no

Normalization supported: yes

Embeddings: 256, 512, 1024

Supported use cases: Semantic similarity searching to find documents (for example, detecting plagiarism), classifying labels into data-based learned representations (for example, categorizing movies into genres), and improving quality and relevance of retrieved or generated search results.

Read the blog

Read the blog

Amazon Titan Multimodal Embeddings

Powers accurate multimodal search and recommendation experiences.

Max tokens: 128

Max images size: 25 MB

Languages: English

Fine-tuning supported: Yes

Embeddings: 1,024 (default), 384, 256

Supported use cases: Search, recommendation, and personalization.

Amazon Titan Image Generator

Generate realistic, studio-quality images using text prompts.

Max characters: 512

Languages: English

Fine-tuning supported: yes

Supported use cases: Text-to-image generation, image editing, and image variations.

Read the blog

Watch Amazon Titan Image Generator demo

Watch watermark detection demo

Watch re:Invent session

Amazon Titan Image Generator v2

Generate photorealistic images, with support for image conditioning, subject consistency, instant customization and background removal

Max characters: 512

Supported input/output image formats: PNG, JPEG

Languages: English

Fine-tuning supported: yes

Supported use-cases: Text-to-image generation, image editing, image variation, image conditioning using a reference image, subject consistency using fine tuning (preserve specific subjects in generated images), and automated background removal.

Read the announcement 

 

Videos

Amazon Titan Text Premier Demo (1:30)
Amazon Titan Image Generator Demo: Image Playground (5:53)
Amazon Titan Image Generator Demo: Watermark Detection (6:11)
AWS re:Invent 2023: Explore Image Generation and Search with FMs on Amazon Bedrock (59:23)
Amazon Titan Text Demo (8:34)
AWS re:Invent 2023: Explore Amazon Titan for Language Tasks (39:00)

Videos

Amazon Titan Text Demo (8:34)
Amazon Titan Image Generator Demo: Image Playground (5:53)
AWS re:Invent 2023: Explore Amazon Titan for Language Tasks (39:00)
AWS re:Invent 2023: Explore Image Generation and Search with FMs on Amazon Bedrock (59:23)

Amazon Titan Image Generator features

Amazon Titan Image Generator empowers content creators with rapid ideation and iteration for highly efficient image generation. Users can guide image creation using reference images, edit existing visuals, remove backgrounds, and generate image variations – all while securely customizing the model to maintain brand style and subject consistency. This powerful tool streamlines workflows, boosts productivity, and brings creative visions to life.

Amazon Titan Image Generator features

Use ControlNet with reference images to generate cohesive, captivating images that preserve key edges and regions.


Input text prompt: a deer in a fairy world, tiny huts on sides, snowy winter season, sun rising, cartoon style

 

Reference image:

reference image

 

 

Output image:

Output image

Automatically remove background from images containing multiple objects without any user input.


Input image:

Input image

 

Output image:

Output image

Effortlessly integrate your subjects in generated visuals by fine tuning the model


Input text prompt: Ron the dog on the moon

 

Reference image for model fine-tuning:

Ron the dog
Ron the dog

 

Output image:

Ron the dog on the moon

Generate favorite subjects in a new scene, transfer styles using reference images or mix styles using multiple reference images without fine tuning  

Input text prompt: a dog wearing a baseball cap


Reference image:

reference image

 

Output image:

output image

Preserve brand's visual identity by precisely controlling the color palette of your generated images - no fine-tuning required.

 

Input text prompt: a jar of salad dressing in a rustic kitchen surrounded by fresh vegetables with studio lighting

 

Input hex code: ['#ff8080', '#ffb280', '#ffe680', '#e5ff80']

 

Color-controlled output image: