MK.IO supports AI workflows for On-Demand video processing.

We handle the work of segmenting video, extracting relevant data, sending the data to AIs, normalizing responses, and then doing something with your generated content. Examples of "something" include embedding generated content into TTML closed captions and secondary-to-denary audio tracks, generating in-band SCTE-35 signals containing generative metadata, transferring descriptive blobs into storage, stashing JSON into the MK.IO data lake.

Right now it seems like the whole world is exploring what is possible with AI — our aim is to make that exploration fast and easy for our customers (you), and to facilitate the rapid rollout of your most successful AI-augmented workflows to your customers.

Whether your workload is live or on-demand, you define and execute your AI workloads the same. So you can focus on building products instead of focusing on building data processing pipelines.

📘

MK.IO uses a third-party AI tool, specifically Azure AI Speech, from Azure AI Services.

Key concepts: Pipelines and Models

An AI Pipeline is a collection of operations that will be performed against your content. Your operations define which Model to execute, the destination outputs for your generated content, and any arguments required by the pipeline (for example, language, helper phrases).

Models provide a menu of the supported operations in MK.IO. They describe the supported input and output formats, and any arguments that may be provided (including where the argument can be modified).

Running an AI workflow

To execute an AI workflow, you just have to attach an AI pipeline to a VOD Transform and run a Job. See Configure transformations.

Supported languages

The table below summarizes the locales available in the AI model for transcription.

*The model is bilingual and also supports English.