Frame.io gives you the power to search however you want, to find whatever you need, in many different ways, from Standard search to our latest visual and semantic search technologies with media intelligence. Read on to discover how to get the most out of your search options!

This feature is available on both web and mobile (iOS) platforms.

Our enhanced search functionality can find any media, folders, and projects across your full account and is designed for creatives and managers who need to efficiently locate, inspect, use and manage digital assets and Frame.io Projects. By leveraging metadata alongside relevance-based ranking, we’re transforming the way users interact with their content, making the search process faster, more intuitive, and more effective.

Begin by clicking on the magnifier icon on the left of any Frame page, clicking the search bar at the top of the Workspace page, or using the shortcut Ctrl/CMD + K.

Beyond just searching general phrases and titles, you can use this search feature to find metadata attached to your media. This includes Status, Keywords, Uploader, Assignee, File Type, Format, and Source Filename.

For example, if you are looking to find all of your media that has the status of Needs Review, typing "Needs Review" into the search bar will show results across your entire account of all media assets with that status listed. If you want to find all your files that are PDFs in your account, just type "PDF" and all your results will be PDF files.

The search feature will display any assets in list form with a preview panel on the right of the results. The preview will show a thumbnail for the asset and file information, including Workspace and Project location, size, file type, and what metadata matches with your search.

You can also combine keywords or metadata into one search by adding a comma (,) to make your searches more accurate.

For example, searching "Audio, .wav, Needs Review" will use three different metadata fields will show the proper results to more accurately find the audio file that needs to be reviewed rather than just searching "Audio". Another example is if you have a media asset with multiple Keywords, searching for as many keywords that apply to your search separated by commas will help you more accurately find the exact file you may be looking for.
​
You can open and view an asset from the search results by double-clicking on the name in the list or clicking the thumbnail preview. You can also click View in Source to open the source location of that media/folder and Copy URL to link to the source page.
​
To exit search, you can click outside of the search box.

Local Search in Frame.io is a quick and simplified search experience that’s linked to your current Project or folder. By using the magnifier icon in the top center of your current Project space, you can search for any keyword and the results of assets in that Project will appear automatically in the main page.

Standard results are particularly helpful when you already know the name of the item they wish to locate. Standard Results derive matches based on name for assets, folders, and projects. For assets, there’s also matches from metadata values for a core set of metadata fields:

Natural Language Search is Frame.io’s newest way to use the power of metadata fields to find more accurate results for the media you’re searching for metadata in your account. It has the ability to handle more complex queries which contain several match requirements across metadata and semantic search.

Instead of searching for project and folder titles, you can use a number of different metadata fields, such as rating, comment count, pages, duration, file size, audio sample rate, and transcription, and many more, to get even more exact search results.

For example, a basic search such as “video uploaded by Jane Doe,” can be improved by using more specific phrases. By adding file type, date uploaded, and keyword metadata’s, the search can become “4k ProRes video uploaded during August by Jane Doe tagged with ‘people’ and status ‘Needs Review’” to get even closer to the results you’re looking for.
​

If a video has a transcription generated or uploaded to it, your searches will also be able to find transcription results. For example, the phrase “Transcript ‘secure’” will show all results with that word in the video’s transcript. Clicking on the matches on the right will open the transcript panel and highlight the results.

Additional metadata fields can also be added for more accuracy as well (ex. “transcript ‘secure’ in the documentary project rated at 3 stars or above”). Search results when selected will bring you to the exact transcript location in the video.

Same can be done with Frame.io comments on your media. Search for a phrase or keyword found in your comments and NLP search will find the results (ex. “Comment 'Asset Manager'”). Search results when selected will bring you to the exact location of the comment.

Semantic search is a visual search using media intelligence that will match words/phrases to find results inside the media itself. Unlike NLP search, semantic search will not be based on metadata and instead be able to recognize keywords to find visual references within the content of any images and videos.

For example, a search that includes “clockface” will show results with any media that has clock or time imagery, regardless of if it is using “clockface” in the title or as a metadata keyword. Clicking into any results will show white lines (subclips) on the timebar highlighting the phrase. You can navigate directly to a specific subclips using the detail panel to the right of the results list.

If you search “wedding footage featuring the groom,” Semantic search will be able to look through all videos that feature “wedding” and “groom” imagery as combined visual references as search results to be more accurate than searching just “wedding” or “groom.”

Results should highlight the most accurate results at the top and give more examples that are close to what is being searched for below. If you search for “black and white images of a person working out,” the first results might show very accurate image results but below will be results of other “black and white” “images” and “working out” examples. Selecting your result will open to the asset at the exact timestamp of the match.

Hybrid searching can consist of combining NLP and Semantic searches in Frame.io, meaning you can search for visual imagery while including metadata fields for a hybrid search. For example, you can combine a phrase like “find videos of a family in the pool AND uploaded in the last 30 days.” The results will show all visual results and Date uploaded metadata field to filter the results to the last 30 days.

There will also be hybrid results in displaying both Standard results and media intelligent results when a search is made. Sometimes a search can be direct to the name of the title in the media, but may also refer to an object found in the frames of the video or in the metadata of the media. If, for example, a search was made for “All my interiors,” the first results will show the Standard results with “Interiors” in the title of the media, folder, project, or keywords. Below that list will show the media intelligent results, which can combine NLP and Semantic results, showing all metadata fields that match interiors or visual representations of interiors, such as bedrooms or lamps.

Media intelligence Results are ideal if the user has a multi-intent query (e.g., matching across three different metadata fields). Media intelligence Results derive matches courtesy of NLP.

Frame.io uses a learning language model, or LLM, to do a smart translation of your conversational speech into meaningful search queries. That being said, there are some barriers to our LLM:

As long as the written text can be understood conversationally, the LLM will be able to provide relevant results.

Media intelligence found in semantic search is scene-based within the frame and will best track anything found prominently in that frame. If an item is too small or far in the background, it may be difficult to find an accurate match.

During the Beta period for NLP and Semantic search, iterations will continue to be made in tuning the ranking of results. For example, most recent results won't necessarily always appear near the top of the list.

The purpose of the Beta period is to collect feedback to help improve this experience for our customers. If you encounter search results that seem unexpected, please provide feedback via the in-app Search feedback button and/or contact support.

Q: Are there any limitations with NLP or semantic search?
A:

Q: Are there any metadata fields not supported by NLP or Semantic search?

A:

Q: Can I search using the Frame.io panel in Premiere Pro?
A: Yes. Premiere will search locally within the current project, while Frame.io searches account-wide and utilizes NLP.

Q: Does semantic search have access to the account’s media assets?

A: Yes. This is how the embeddings are created. We offer you the ability to opt out. Just contact support to let us know and your assets won’t be processed for semantic search.

​Q: Does NLP have access to the account’s media assets?

A: No. NLP is only used to parse the text query entered by the user. It has no awareness of the content against which the final query will be run.
​

Q: Will customer assets be used to train/reinforce our AI model?

A: No. In line with Adobe’s industry-leading ethical posture, we do not use customers’ data to train or augment our internal model

Q: The new feature will generate references to visual objects. What is that and where will the associated metadata be stored?

A: The references are sets of decimal numbers that only Adobe can decode. The references are stored in our secure internal database. No decoding takes place other than when the customer searches in their account.

Q: Does semantic search support finding specific people in pictures and videos using facial recognition?

A: No. This is functionality under consideration.