At XL8, we specialize in state-of-the-art AI-powered media localization tools and work with some of the leading content producers in the media industry. We are often entrusted with highly sensitive data and content, and the imperative of maintaining confidentiality over media files and transcripts of unreleased films and TV shows cannot be overstated. Thus, data and content security are not just priorities but foundational pillars of our operations. Our attainment of the SOC 2 and Trusted Partner Network (TPN) certifications is a small testament to our commitment to the most rigorous security standards.
Cloud-Based Solution
While our cloud-based solutions offer convenience, scalability, and the highest quality results, there is no getting around the fact that the sensitive content must be uploaded to our infrastructure for processing. We are happy to discuss alternative methods to resolve any lingering security concerns. For example, some clients may opt to only upload audio tracks for transcription. This will limit our ability to produce the optimized time codes based on the shot change frames, or provide our enhanced speaker diarization feature. However, depending on the use cases, it can be a reasonable compromise. Some clients choose to encrypt the media files before uploading them to our secure storage, and share the encryption key with us through a separate channel.
Why On-Premise Solution?
Guaranteeing Content Security
The foremost priority of the on-premise solution is to ensure that the contents never leave the client's infrastructure. The simplest way to achieve this is to cut off the server from the Internet. However, it is still necessary for the server to communicate with XL8's license server to validate the license and track usage. To do this without any doubt on the content security, our on-premise solution requires allowing only HTTPS GET requests to a single endpoint. It sends a small payload, containing license information and process sizes to the license server on start-up and around each operation. Outgoing traffic can be restricted to our license server's IP addresses. To be extra sure, you can set up a HTTP proxy to only allow GET requests to the specified endpoint, and monitor the traffic.
Seamless Integration: On-Premise Docker
Thanks to the widespread adoption of Docker, streamlining the integration process has become a straightforward process. By developing a Docker image that can perform translation locally, clients can quickly deploy the solution and expect our application to run consistently diverse environments thanks to Docker's containerization technology. Moreover, any updates can be easily distributed by creating an updated image and pushing it to the Docker image registry.
The basic XL8 on-premise server is available as an all-in-one Docker image that contains a simple API server and a translation model for a specific language pair. Starting the translation server is as simple as pulling the image from our Docker registry, and running it on any of your servers. The REST API provided by the container allows you to request translations and receive the results programmatically. A simple web UI is also provided for simple translation tasks. For scalability, it allows launching as many containers as needed across servers to process a large amount of translations in parallel.
However, this approach has some downsides as well. A separate Docker image is needed for every language pair you need, and each image is rather large. As a machine translation engine specializing in media localization, we often prioritize the translation quality over model size, hardware requirements, or translation speed. Each container can take up rather a large amount of resources, and memory usage may vary between language pairs. So if you have translation needs across multiple language pairs with varying usage amounts, it can be tricky to allocate resources and scale them accordingly.
Scalable On-Premise Solution on Your Cloud
For clients seeking higher levels of scalability and flexibility, we also offer an advanced solution that leverages scalability of the cloud provider, by deploying our media localization system directly on your AWS account. We provide two types of Docker images hosted on Amazon Elastic Container Registry: one for the API server and another for translation process. The API server image can be launched in any environment, with an XL8 license file and a small file containing your AWS configuration. Upon receiving a translation request, it initiates a translation container on your AWS Fargate cluster. Every model file, encrypted with your license key, will be replicated to your S3 storage, so a single translation Docker image can translate across any language pair by downloading the necessary model files from your S3 storage.
This solution does require a more complex setup process, including creating ECS task definitions and some S3 buckets, and replicating model files from our S3 bucket to yours, and configuring an IAM role with permissions tailored to your security requirements. While the initial setup may be intricate, the resulting environment will be unparalleled in terms of scalability, flexibility, and ease of maintenance. It scales up and down automatically according to your translation needs, and all operations are conducted within the confines of your AWS account, ensuring a high level of security and control.
AI Solutions Tailored to Your Security Needs
As an AI start-up in the media industry, our mission is to optimize and accelerate the localization process while meeting your security expectations. Offering an on-premise solution represents just a fraction of our comprehensive efforts to meet diverse needs. If you have considered adopting AI into your localization workflows but have hesitated due to content security concerns, we invite you to start a discussion with us. We are committed to identifying a tailored approach that delivers the highest quality of results possible within any restriction you may have.
