What challenges did Fotor face before AI implementation?

Reduce inference latency for high concurrency across a global user base. Improve automated asset tagging and content moderation. Integrate generative AI into existing image editing and design workflows.

What was the impact of the AI solution?

Processing time per request dropped from 10-20 seconds to 7-8 seconds. Fotor handles 300 concurrent requests per second. Hundreds of generative AI features were launched. Daily active users increased tenfold and user satisfaction rose 20%.

Rank#1,773of 3,107

Source

Fotor: Generative AI image editing using Amazon Bedrock, SageMaker, and Rekognition

Use case typeContent generationUpdated Jul 1, 2024

Fotor uses Amazon Bedrock, Amazon SageMaker, Amazon Rekognition, Amazon SQS, Amazon SNS, and Amazon EC2 to power generative AI image editing and design tools for 600 million users worldwide. The solution reduces inference latency, automates image tagging, moderates user-generated content, and supports custom image and video model development.

Organization: Fotor
Industry: Consumer & Food
Location: China
Published: July 2024

Reported outcomes

+20%

user satisfactionCustomer experience

−50%inference time per request300 requests per secondconcurrent requests+1,000%daily active users

Strategic outcomes

New product / capabilityLaunched hundreds of generative AI featuresNew business modelDrove revenue growthCost efficiencyReduced manual effort in tagging and moderation

Catalog median for customer experience deployments: +69% across 99 reported metrics. Compare benchmarks →

Primary read

Use case focus

Showing 3 of 3

1Content generation
2Content review and revision
3Customer personalization

Reduce inference latency for high concurrency across a global user base.
Improve automated asset tagging and content moderation.
Integrate generative AI into existing image editing and design workflows.

Amazon SageMaker asynchronous inference with SQS/SNS and EC2 scheduling to process high concurrency requests.
Amazon Bedrock LLMs for semantic expansion and image/visual annotation.
Amazon Rekognition for content moderation confidence scores.
Amazon SageMaker model training to develop proprietary image and video models.

Technologies

Amazon Bedrock Amazon SageMaker Amazon Rekognition Amazon SQS Amazon SNS Amazon EC2

Processing time per request dropped from 10-20 seconds to 7-8 seconds.
Fotor handles 300 concurrent requests per second.
Hundreds of generative AI features were launched.
Daily active users increased tenfold and user satisfaction rose 20%.

Architecture

Fotor runs its services across multiple AWS regions and uses Amazon SageMaker asynchronous inference integrated with Amazon SQS, Amazon SNS, and Amazon EC2 scheduling to process high-concurrency requests. It uses Amazon Bedrock for LLM-based image annotation and semantic expansion, Amazon Rekognition for moderation confidence scores, and Amazon SageMaker training for proprietary image and video model development.

Sources & evidence1

Groundedness: 4/5Type: Customer StoryPublished: Jul 1, 2024Publisher: AWSEvidence: PrimaryConfidence: High

Primary source

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Customer evidence