AI has an important emerging role in managing your media. It can help you to streamline processes by removing repetitive manual tasks, freeing up your human resources to do tasks where human judgment is beneficial.
Curator’s AI integrations focus on exploiting types of AI that are already reliable so that you have more time to concentrate on creating great content.
What can Curator add?
AI can be a valuable source of automated metadata. However, an increase in volume of content can mean that you’re deluged with data you don’t have the resources to make sense of. Curator can help bring meaning to this data by presenting it clearly with provenance and ranking, so you understand where it’s from and how valuable it is.
Speech-to-text
In a recent IFTA survey, speech-to-text was identified as the type of AI that is currently most useful in daily work.[1] Curator integrates with AWS Transcribe and Azure Media Services to provide speech-to-text functionality within the Curator applications.
Benefits of using speech-to-text with Curator are:
- Curator transforms speech-to-text into readable sentences that can be used as captions.
- Curator helps you to make the most of your investment by making speech-to-text results easily searchable.
Use cases include:
- Help media managers to organize content by automatically creating accurate, searchable metadata about each scene using the caption text produced.
- Automating a base level of captioning to help content producers save time when captioning their content.
- Help executives to save money by easily and affordably meeting the legal need for captioning under the Americans with Disabilities Act (ADA).
Object recognition
Object recognition is another emerging technology which can be useful if deployed in the right way. But AI-created metadata can create more challenges because of the sheer quantity of tags it produces – 60 minutes of content can produce 214,000 labels, even if you’re filtering out all content below 50% confidence rating!
Our integration with Amazon Rekognition provides another way of tagging every scene with accurate metadata, while avoiding being deluged.
Here are some benefits of using object recognition with Curator:
- Controlled vocabularies allow you to define which terms you want to include as searchable data in Curator. This allows you to include high-value terms and filter out low-value to avoid data deluge. For example, in the below image you’re likely to be most interested in the lions, not the grass/mammals/nature tags that might also appear.
- Confidence – Curator allows you to globally filter out low confidence data and change the threshold for filtering on a per-search basis.
- Label clustering thresholds let you group adjacent labels into subclips. You can also set maximum duration thresholds to automate segment logging.
- Create an AI profile to reflect your needs by choosing the type of data you want to include: content labels, moderation and/or celebrity detection.
- Keep the raw data so it can be re-processed anytime if you tweak the processing options, without needing to pay for the video to be re-analyzed by the AI provider.
Use cases include:
- Using object recognition for high value insights while filitering out noisy metadata – great for media managers.
- Allowing editors to access object recognition labels without leaving their editing tools and losing their flow - great for creative directors who want to integrate AI into their editing workflow.
Notifying you of scenes which you may need to edit out for compliance purposes – cutting out dull work and allowing creatives to spend more time producing valuable content.