Generative AI Applications with Amazon Nova PDF Free Download
1 / 50/50
100%
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
ZURICH | SEPTEMBER 11 2025
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Generative AI Applications with
Amazon Nova
Andrzej Mizerkiewicz
GAI- 3 0 6
Sr Spec Solutions Architect AI/ML
Amazon Web Services
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova
Foundation
Models
State-of-the-art foundation models that
deliver frontier intelligence and industry-
leading price performance
Amazon Nova
Lite
Amazon Nova
Premier
Amazon Nova
Pro
Amazon Nova
Micro
UNDERSTANDING MODELS
Amazon Nova
Reel
Amazon Nova
Canvas
CREATIVE CONTENT GENERATION MODELS
SPEECH-TO-SPEECH MODEL
Amazon Nova
Sonic
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova:
Understanding Models
Amazon Nova Micro, Amazon Nova Lite,
Amazon Nova Pro, and Amazon Nova Premier
4
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova
Lite & Pro Only
Amazon Nova
Micro | Lite | Pro | Premier
Amazon Nova
Micro, Lite, & Pro
Generate
text and code
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Power agentic
workloads
Decompose complex problems
into executable steps
Integration into external systems,
including multi-step function calling
UI actuation, browser
action automation
Strands Agents SDK
Integrated
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Integration with
Amazon Bedrock
Bedrock Knowledge Bases
Bedrock text and vision fine-tuning
Bedrock Agents
Bedrock Guardrails
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova Understanding Models
Amazon
Nova Micro
Amazon
Nova Lite
Amazon
Nova Pro
Amazon
Nova Premier
Availability
GA GA GA GA
Context window
128K 300K 300K 1M
Languages
200+ languages 200+ languages 200+ languages 200+ languages
Modalities supported
Text input;
text output
Text, image, video
input; text output
Text, image, video
input; text output
Text, image, video
input; text output
Fine
-tuning Yes Yes Yes No (Teacher for
distillation)
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Customize across all stages of model training
Customization Technique Training
Amazon Bedrock Amazon SageMaker AI
Supervised Fine
-Tuning
•Parameter-efficient fine-tuning ✓✓
•Full fine-tuning ✓
Direct Preference Optimization (DPO)
•
Parameter-efficient DPO ✓
•
Full model DPO ✓
Proximal Policy Optimization (PPO)
✓
Continued Pre
-Training ✓
Knowledge Distillation
✓ ✓
N E W
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova Price-Performance Tiers
Improve intelligence at your desired price point with fine-tuning and model distillation through Amazon Bedrock
I N T E L L I G E N C E
PRICE
Micro
Lite
Pro
Premier
Fine tuning
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon
Nova Micro
Amazon
Nova Lite
Amazon
Nova Pro/Amazon
Nova Premier
•
Summarization
•
Simple Function Calling
•
Support intent Classification
•
Entity extraction
•
Documents with charts and
images
•
Extracting data from images
•
Planning for multi-step
workflows
•
QA against business data
•
Extracting metadata from videos
•
Mutli-image visual reasoning
•
Complex reasoning and plan
orchestration
•
Generating complex SQL queries
Amazon Nova – what is the right model for me ?
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Fast output generation is key for many real- world use cases
Think Fast, Think Nova Amazon Nova models are designed for
low latency understanding applications
Output Tokens per Second Time to First Token (seconds)
Source: Artificial Analysis, 8/26/2025
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Price per 1M Tokens
Nova Pricing
Source: Artificial Analysis, 8/26/2025
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Nova Long Context
Designed for high accuracy on long documents
Needle Position (in %)
Amazon Nova
Micro 128k
Amazon Nova
Lite 300k
Amazon Nova
Pro 300k
Amazon Nova
Pro 5M
Coming Soon!
Amazon Nova's Context
High Accuracy across different context lengths
Amazon Nova
Pro 1M
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Native Video Understanding
Input Prompt: Describe the video
Input Video:
Video
Prompt The video depicts a football game in progress,
featuring two teams on a well-marked green
field. Initially, the teams are positioned in a
classic formation, with players from both sides
facing each other, ready for the play. The
players are equipped with helmets and jerseys,
indicating a formal match. As the video
progresses, the play commences with the
offensive team executing a pass, leading to a
tackle by the defensive team. The action is
intense, showcasing the physicality and
strategy inherent in football. The camera
captures the play from an aerial perspective,
providing a clear view of the field and the
players' movements.
Output
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Handwriting understanding
Input Prompt: Transform this whiteboard into JSON.
Identify use case and features that can/cannot enable
each use case.
Image
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Handwriting understanding
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Diagram understanding
Input Prompt 1
You are an engineer, review the diagram and
tell me what the process flow is about in detail
and where it would be used
Input Prompt 2
How would you improve this process?
Input Image
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova
State-of-the-art foundation models
that deliver frontier intelligence and
industry leading price performance
Amazon Nova
Reel
Amazon Nova
Canvas
CREATIVE CONTENT GENERATION MODELS
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova
Canvas
Image generation model
Amazon Nova
Reel
Video generation model
Amazon Nova
Amazon Nova creative content
generation models are state-of-the-art
image and video generation models
with built-in controls for security,
safety, and responsible AI.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova
Canvas
Amazon Nova
Reel
OUTPUT
INPUT
Text Input
Image Input
Image
Video
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova:
Creative Content
Generation Models
Amazon
Nova Reel
Amazon
Nova Canvas
Availability
GA GA
Input characters
512 1024
Languages
EN EN
Modalities supported
Text, image input;
video output
Text, image input;
image output
Duration
6 seconds
(2 minutes multi-shot) N/A
Fine
-tuning COMING SOON YES
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova
Canvas
State-of-the-art image
generation model
Enables creation and editing of studio-
quality images for advertising, marketing,
e-commerce, and other domains
E.g., a natural exfoliating body scrub with a vibrant
colorful splash of mint and cucumber juice in a
bathroom with a beautiful natural view.
Text to image
Generate high-quality images from natural language prompt
Automatic editing
Edit an image with a text prompt (no image mask)
Inpainting/outpainting
Remove/replace an object; Extend boundaries of an image
Image variations/instant customization
Generate variations with up to 5 images as input
Image conditioning, background removal
ControlNet (canny, segmentation);color palette for brand colors
Watermarking and content moderation
Traceability, detection and removal of harmful content
Coalition for Content Provenance and Authenticity (C2PA)
Add metadata by default to generated images
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
A dinosaur
sitting in a
tea cup
Text to image
Generate images from text prompts
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Automatic editing - inpainting
Mask-free editing
Mask prompt: “lake”
Text prompt: “field with fence”
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Automatic editing - outpainting
Generate different backgrounds
Mask prompt: “coffee maker”
Text prompt: “a coffee maker in a sparse
stylish kitchen, plate of pastries next to
the coffee maker, a single cup of coffee”
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Instant customization
Transfer styles using 1-5 reference images
“a red haired boy sits in a school classroom
looking bored, illustrated story”
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Image conditioning
Control pose, camera angle, and more
Output - Canny Edge
Output - Segmentation
“3d animated film style, a woman
with a crazy blond hair style,
wearing a green sequin dress”
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Color Palette
Control the color palette of generated images
“digital painting of a girl, dreamy
and ethereal”
["#81FC81", "#386739", "#C9D688",
"#FFFFFF"]
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Background removal
Remove background from images
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Virtual try-on
Create immersive shopping experiences by visualizing products in context
Reference image
Source image Output image
•Integrate directly into UX
•Naturally drapes clothing and accessories
•Intelligently places furniture and decor
•Preserves details (logos and graphics)
•Enhances purchasing decisions
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Style options available in Nova Canvas
Artistic styles that will consistently produce an image using that style, without specifying
within the prompt
8 supported styles
•3D animated family film
•Design sketch
•Flat vector illustration
•Graphic novel illustration
•Maximalism
•Midcentury retro
•Photorealism
•Soft digital painting.
“a happy elephant”
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
34
Amazon Nova:
Creative Content Generation
Amazon Nova Reels, Amazon Nova Canvas
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova
Reel
Video generation model
Enables creation of short videos through
simple prompting via input images or a
natural language prompt
E.g., perfume bottle with jasmine and citrus
themes and a calming, natural setting. There are
trees in the background, the leaves are gently
fluttering in the breeze
Text to video
Generate videos from natural language prompt
Image to video
Provide a text and an image to generate video
Camera motion control
Control camera motion with text prompting
Built-in controls for safe and responsible AI
Watermarking, content moderation
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
“Cinematic dolly shot of a juicy
cheeseburger with melting cheese,
fries, and a condensation-covered cola
on a worn diner table. Natural lighting,
visible steam and droplets. 4k,
photorealistic, shallow depth of field”
Text to video
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
“Closeup of a large seashell in the
sand, gentle waves flow around the
shell. Camera zoom in.”
Text to video
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
“Dolly forward”
Image to video
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
41
Amazon Nova:
Speech-2-Speech
Amazon Nova Sonic
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
INPUT
Speech
Amazon Nova
Sonic
Speech
OUTPUT
Text
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon
Nova Sonic
Amazon
Nova Sonic
Availability
GA
Context window
300 K
Modalities supported
Input: Speech
Output: Speech and text
Languages
English, French, Italian, German, Spanish,
with more languages coming soon!
Other Features
Tool use for function calling, knowledge grounding
Real
-time streaming Bidirectional streaming Bedrock API in IAD, NRT and
ARN regions
State-of-the-art speech-to-
speech model that delivers
real-time, human-like voice-
first conversations.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Conversational AI
Features
Best-in-class speech recognition: Accurately recognizes streaming speech
across accents with robustness to background noise.
Low user perceived latency: Bidirectional streaming speech I/O with low
user perceived latency of ~1.09 sec @P50*.
Fluid dialog handling and natural turn taking: Handles user’s speech
interruptions, detects non-verbal cues (e.g. laughter, grunts, inter-sentential
pause and hesitation) to enable human like turn taking in dialogs.
Adaptive speech responses: Adapts speech response to user’s tone and
sentiment.
Emergent properties: Supports tool calling, RAG based info retrieval,
Responsible AI and privacy compliance.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Amazon Nova Sonic
Key attributes
✓Input modalities: Speech
✓Output modalities: Speech and text
✓Context length: 300 K Tokens
✓Languages: English and Spanish
✓Latency: Time to first audio (TTFA) (p50) of 1.09 sec
✓Price/K Tok:
•Speech: Input: $0.0034; output: $0.0136
•Text*: Input: $0.00006; output: $0.00024
*The text tokens input and output pricing applies to specific use cases
such as speech-to-text transcription, tool calls for task completion or
knowledge grounding, adding conversation history to the session etc.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Language learning for
non-native speaker
See Amazon Nova Sonic In Action – Example #1
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Voice-enabled
business assistant
See Amazon Nova Sonic In Action – Example #2
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Customer service
call automation
See Amazon Nova Sonic In Action – Example #3
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
nova.amazon.com
A new website designed to easily
explore the latest Amazon Nova
foundation models
To start building gen AI applications with
Amazon Nova, you must access the models
through an API using Amazon Bedrock.
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Thank you! Please complete the session
survey in the mobile app
Thank you
Andrzej Mizerkiewicz
mizerkie@amazon.pl
