May was about turning the library into a workspace. The headline features all point the same direction: structured metadata your team defines, conversations that live on the asset, and the precision tools reviewers need to catch the thing that matters. Here is the tour.
Custom metadata fields, defined by you
Ceivo's AI gives every asset a rich description, but every media operation also tracks its own facts: air date, clearance status, rights window, delivery spec, market. May added custom fields, a catalog of typed metadata your admins define once and your team fills in per asset.
A field has a type (text, number, true/false, or date) and applies to the file types you choose, so a "Cleared for Broadcast" checkbox shows up on video while an "Air Date" calendar field shows up where it belongs. Fields work on folders as well as files, which means a whole title or campaign can carry its own structured attributes, not just the clips inside it. Because the values are typed rather than free-text tags, downstream agents and integrations can read and reason over them reliably.
Comments and reactions, right on the file
Conversations about an asset used to happen in email, chat, and spreadsheets, everywhere except the asset itself. May brought a threaded comment system to every file inside Ceivo. Comments support replies, can be anchored to a timecode on video and audio, can be marked resolved when an issue is closed, and now carry emoji reactions so a reviewer can sign off without adding another line to the thread. The same reactions are available on share comments, so the people you send work to can react in the browser without an account.
This is the connective tissue behind the brand-approval work we shipped this month. The threaded, version-anchored conversation that powers Ceivo Atrium is the same comment model, now available on every file in the library.
Region markers on images
Markers used to be a video idea: a name and a span of time. As of May, images get spatial markers, a named, described box drawn over a specific region of the frame. A reviewer can point at the exact corner where a logo crosses the safe-area margin, or the spot where a key-art element breaks brand standards, instead of describing it in prose. It is the difference between "the logo placement is off" and a box on the pixels, and it is what lets AI agents return findings a human can act on at a glance.
See exactly what changed, side by side
When you have an original and a re-encode, or a master and an altered copy, the question is always the same: what actually changed, and where? May added visual file comparison, which walks two files frame by frame and reports how far apart they are across the timeline. The result is a difference profile that points you straight at the moments that diverge, so you can confirm a clean transcode or zero in on an unexpected alteration without scrubbing both files end to end.
File versioning that holds the line
Assets evolve: a master gets a revision, a config gets a new cut, a deliverable gets a fix. May added a proper parent and child relationship between files, so a new version can point back to the one it came from and the full chain travels with the work. Each version stays immutable, the lineage is explicit rather than implied by filename, and you can walk from any asset to the versions derived from it. It is the backbone for revisions in review workflows and for any place you need provenance to survive a change.
Top 10 notable improvements
Beyond the headline features, ten smaller items worth calling out:
- Hidden folders: hide folders you do not need from your own view without changing anything for the rest of the team.
- Folder custom fields: the same typed metadata catalog now applies to folders, not just files.
- Reference links on markers: attach a source link to a marker so a finding deep-links to the asset or guideline it was checked against.
- Walk a file's versions: list every file derived from a given asset in one step.
- Detach a version: break a parent relationship cleanly when an asset is no longer a derivative.
- Resolve and reopen comments: toggle a comment's resolved state as issues open and close.
- Timecoded comments: anchor a note to the exact moment on a video or audio file.
- Archive agent runs: clear finished agent runs out of the default view and bring them back when you need them.
- Reactions on shared work: recipients can react to comments on a share link directly in the browser.
- Cleaner pagination: list endpoints now accept a zero-based offset, a small fix that makes integrations more predictable.
Why this month mattered
For two years the case for Ceivo has been search and analysis: point the AI at an archive and find anything. May broadened that into collaboration and governance. Custom fields let teams encode their own operational truth alongside the AI's. Comments and reactions put the conversation on the asset. Region markers, comparison, and versioning give reviewers the precision and the paper trail that real approval workflows demand. Together they are what make the agent-driven review surfaces we have been shipping, from the spring agent catalog to Atrium, feel like a product rather than a pipeline.
What's next
Want early access to something you saw above, or curious how custom fields or review workflows could fit your operation? Reach out and we will take it from there.