It is easy to be skeptical of a features roundup. Most of them read like a changelog with adjectives added, a list of things that shipped without much sense of why any of them matter to the actual job of running a studio. This is not that. Every feature below earns its place because it slots into a specific moment in the work you already do, from the moment you dump a card to the weeks after a client has their files. Here is where each one fits.
The moment you sit down with a full card
Every shoot ends the same way: hundreds or thousands of files sitting in a folder, and the immediate need to find the handful that matter without scrubbing through everything by eye. This is the stage where AI photo tagging does its quiet work. As photos upload, a vision model reads each one and populates searchable metadata automatically, subjects, settings, notable details, the sort of thing you would otherwise have to keyword by hand or simply never bother tagging at all. Paired with it is a newer addition, an AI pick score computed for every photo, a quality signal the same vision pass generates and quietly uses to boost your strongest shots when you search. You are not doing anything differently to get either of these; they happen in the background the moment photos land.
The payoff shows up the next time you need something specific. Unified search reaches across every gallery and video collection a company has ever delivered, matching against tags, descriptions, AI-generated metadata, and even video transcripts in one query. Ask for a client's name and a keyword, and it surfaces the right gallery instead of you remembering which of forty folders it lives in. For a studio with years of delivered work sitting in the account, this turns a five-minute hunt into a five-second one.
Fixing subtitles without leaving the browser
Every AI feature further down this list, search, summaries, repurposing, Milo, is only as good as the transcript underneath it, and auto-generated subtitles are never perfect on the first pass. A misheard name or a garbled technical term used to mean re-exporting an .srt file in a separate app and re-uploading it. Now the Subtitles tab on a video's admin page is a real editor: click any cue's text and it pauses the video, seeks to that moment, and drops the row into an inline text box so you can fix it while watching exactly what was said. Confirmed edits are marked with an amber border so you can see what is staged but not yet saved, and Prev and Next buttons step you through cue by cue for a fast, sequential pass.
Clicking Save Subtitles sends every cue back in one request, overwrites the underlying .srt file, and automatically re-parses the transcript, which re-queues anything downstream that depends on it, the AI summary, repurposing suggestions, and search indexing all pick up your corrections without you touching them separately. It turns subtitle cleanup from a side trip into a five-minute pass you do right where you are already reviewing the cut.
Reviewing your own edit before the client ever sees it
Video work has a step photo work mostly skips: sitting with your own footage and figuring out what you actually have. This is where Ask Milo earns a place in the workflow. Milo is a chat assistant grounded specifically in the video you are viewing, its subtitle transcript if one exists, the client's business summary if you have generated one, and every timestamped comment left on any version of that video. Ask it what the video's main topic is, or to summarize the key moments for a client recap, and it answers from that context rather than guessing. It will say plainly when a question cannot be answered from what is available, which matters more than it sounds; a tool that admits its limits is one you can actually trust mid-edit. A transcript you have just cleaned up in the Subtitles tab makes Milo's answers sharper too, since it is reading the same corrected text.
Milo lives on the admin video detail page, not the client-facing view, so it is squarely a tool for you during review, not something a client interacts with. If you have ever re-watched forty minutes of raw footage just to confirm whether a specific moment made it into the cut, this is the feature that saves that re-watch.
Protecting a draft while it is still a draft
Between your edit and the client's approval sits a vulnerable window, a cut that is good enough to share for feedback but not yet the file you want circulating freely. Video watermarking handles this without extra steps: apply a watermark at the collection level or override it on an individual video, and the client sees a watermarked version during review while the clean original stays withheld until approval and final delivery. It is the video equivalent of a proof, letting the client judge the cut honestly without the draft becoming the deliverable by accident.
Getting a decision instead of silence
Once a client can see the work, the job becomes getting a real answer out of them, on both photos and video. On the photo side, this is what we mean by proofing: clients star their favorites and comment directly on individual images, and you review everything from a dedicated panel rather than an email thread. We cover that whole loop in depth in our piece on what photo proofing actually is and the full proofing workflow, so we will not repeat it here. On the video side, the equivalent is versioning and timestamped comments: upload a revised cut, collect comments pinned to the exact moment they refer to, and let the client approve or request changes without any of it turning into a scattered thread across email and text.
Turning the finished project into more than one deliverable
The work does not have to stop at the primary file. Video repurposing looks at a video's transcript, the same one you may have just cleaned up in the Subtitles tab, and suggests segments worth cutting into short, social-ready clips, giving you a starting point for the highlight reel or teaser you would otherwise have to scrub for by hand.
On the photo side, any gallery can become a photo slideshow, a real music-video-style file built automatically rather than assembled by hand. Pick how long each photo stays on screen, choose from a range of transitions, add an optional soundtrack with a preview before you commit to it, and bookend the whole thing with an opening and closing slide if you want a title card or a thank-you screen. It builds in the background on the same throttled queue used for transcoding, so a large gallery does not bog anything down, and the finished file shows up with play and download buttons once it is ready. You can generate one for any gallery as the admin, and if you want to hand that ability to a client, it is a one-time, per-client toggle rather than something you configure job by job. Both repurposing and slideshows exist for the same reason: the raw deliverable you already made is often one small step from becoming a second, shareable piece of content, and neither one makes you build that second piece from scratch.
Knowing what happened after you hit send
Delivery has always had a blind spot: once the link goes out, you generally have no idea what the client actually did with it. The client activity log closes that gap. Every view and download a client or guest makes is logged, with the timestamp, and surfaced both as a dedicated activity page and as at-a-glance counts on the client's own info card. You no longer have to guess whether a quiet client has actually opened their gallery or is still sitting on the notification email; you can see it. It is a small feature with an outsized effect on how confidently you can follow up.
Why this matters as a set, not a list
None of these features exist to look impressive on their own. Tagging and pick scores make your own archive usable months later. A cleaner subtitle editor and Milo make the review stage less risky and less tedious, and they reinforce each other since one feeds the other. Watermarking keeps a draft a draft. Proofing and versioning turn client silence into an actual decision. Repurposing and slideshows squeeze a second deliverable out of work you already did. Activity logs tell you what happened after you stopped watching. Strung together, they cover the entire arc from a full memory card to a client who has quietly downloaded their files weeks later, which is really the only test a new feature should have to pass: does it make an actual stage of the job easier, not just the demo.