Find Any Moment: How Aura Searches, Questions and Clips Your Footage (2026)
Find Any Moment is the part of the Aura Suite that removes the slowest step in editing: finding things. Describe a moment the way you would say it, in any language, and Aura jumps to the exact seconds in your library, or ask the footage a question and it points you to the answer. Type "make a highlight reel" and it scores every moment, keeps the peaks and stitches them into one clip ready to export.
Anyone who has cut a two-hour recording knows where the time goes. It is not the cutting. It is scrubbing for the line you remember, the laugh you need or the take where the demo finally worked. Loopdesk treats that as a search problem: every upload is tagged automatically, so you search the library the way you would describe a moment to a colleague, and the result lands on the timeline with one click.
This post covers the three Aura capabilities that do that work and sticks to what happens on screen in the Loopdesk editor. For how AI systems read video at all, see our explainer on AI video understanding.
Product details current as of October 27, 2026.
What does "find any moment" mean in Loopdesk?
It means you reach any second of any clip in your library by describing it: no scrubbing, no transcript search, no logging first. Find Any Moment is part of the Aura Suite, and the homepage puts it this way: "Describe a moment the way you'd say it, in any language, and Aura jumps to the exact seconds. Every upload is auto-tagged, so nothing gets lost."
Three capabilities sit behind that promise.
| Capability | What you do | What Aura does |
|---|---|---|
| Find Any Moment | Describe a moment in the search bar above your library | Scans the library and returns the matching clips with exact timestamps; one click adds a moment to the timeline |
| Ask Your Footage | Ask what is covered, who says what or where the best take starts | Answers from the footage and points you to the moment |
| Highlight Reels | Type one prompt, such as "make a highlight reel" | Scores every moment, proposes a plan you accept or deny, then keeps the peaks, cuts the rest and stitches one ready-to-post reel |
File-name search finds files. Transcript search finds words, if you remember the exact phrase. Neither finds "the moment everyone starts laughing", because nobody says "laughing" while they laugh. Semantic search matches the meaning of your description, not its wording; for how AI derives meaning from frames and audio, see the glossary entry on visual understanding.
How does semantic search work in Loopdesk?
From the editor's side, four steps.
- Upload. Drop footage in, or pull it in with cloud media import. Loopdesk takes all major video and audio formats, including MP4, MOV, MKV, AVI, WAV and MP3, and tags every upload automatically as it lands.
- Describe the moment. Type it into the search bar above your library the way you would say it out loud: "the moment everyone starts laughing".
- Pick a result. Aura scans the library and returns the clips that match, each with the exact seconds where the moment happens.
- Add it to the timeline. Click a result and that moment lands on the timeline, where you trim it by hand or hand the next step to Aura.
Example searches and what comes back:
| What you type | What comes back |
|---|---|
| "the moment everyone starts laughing" | The clips that contain it, each with the exact seconds where the laugh starts, ready to add to the timeline |
| "Find the part where the demo crashes" | The clip and the timestamp where it happens, so you can cut around it or keep it for the bloopers |
| "where she explains the pricing change" | The seconds where that explanation starts, so you can check the wording before you clip it |
| "el momento en que todos empiezan a reír" | The same kind of result. Search takes your description in any language; there is no need to translate it first |
Two details matter. The search runs across the whole library, so a result can come from a file you had forgotten about; that is what "nothing gets lost" means. And the result is a position, not a file: you land on the seconds you asked for, not at the head of an hour-long recording.
Once the moment is on the timeline, the usual Aura prompts apply to it: "Make the intro punchier", "Zoom in on every key point", "Add bold captions". For the last trim by hand, the timeline is built for JKL scrubbing and frame-accurate cuts, covered in our post on Loopdesk's keyboard-first timeline.
How do you ask your footage a question?
Ask Your Footage is the second capability: your footage can answer questions now, and Aura points you to the moment that backs the answer.
In the demo, over a clip with two people talking, you type "What's the main topic of this clip?" and Aura answers: "This clip discusses product-market fit for SaaS startups, featuring 2 speakers." Two facts in one answer. The topic tells you whether the clip is on brief. The speaker count comes from speaker detection, the same capability behind multicam switching and the prompt "Switch cameras on each speaker". Our guide to speaker diarization vs transcription covers what speaker detection can and cannot tell you.
Three kinds of question, each replacing a job you would otherwise do by watching:
- What is covered. The main topic, and whether a subject comes up at all. Use it to confirm a clip belongs in the piece before you start cutting it.
- Who says what. Which person makes a point, and where. Use it before you caption a quote with a name.
- Where the best take starts. When you recorded a line five times, ask for the clean one instead of listening to all five.
How do found moments become a highlight reel?
Highlight Reels is the third capability, and it inverts the first two. Search finds the moment you already know you want. The reel prompt finds the ones you do not.
The demo runs on a single 4:32 take.
- Type the prompt. With the take open, you type "make a highlight reel".
- Read the plan card. Aura watches the footage and proposes a plan before it cuts anything: "4 peak moments", "4:32 → 0:42", and the rule it intends to use, "Keep the four highest-energy moments and stitch them into one clip". The card has two buttons, Accept and Deny.
- Accept or Deny. Accept, and Aura cuts the four moments, joins them and stitches the reel: 0:42, ready to export. If the plan is wrong, choose Deny and nothing is cut.
The plan card is the important part. You see the count of moments, the before and after duration and the scoring rule before any cut is made. That is the agentic pattern: plan, approve, execute. Our explainer on agentic video editing covers it, and the glossary covers highlight clip generation as a category.
The reel is a timeline, not a locked file. "Add bold captions" gives word-by-word animated captions in 108 languages, and a vertical 9:16 version keeps the speaker centered. You can also ask for three separate short clips from the best moments instead of one reel. Export is 1080p or 4K with no watermark on any plan, and cloud renders run at 12× realtime, so a full 4K export lands in seconds, not hours. The finished file sits in the Content Library, the Distribution Calendar schedules it, and Performance Analytics reports views, watch time, subscribers gained and engagement.
Where does this save the most time?
Anywhere the recording is long and the usable part is short.
| Format | The slow part today | What to type | Guide |
|---|---|---|---|
| Podcasts | Finding the quotable lines in an hour of talk, and who said them | "the moment everyone starts laughing"; a who-says-what question; "Switch cameras on each speaker" | How to edit a video podcast |
| Webinars | Pulling ten clip-worthy moments out of slides, demos and Q&A | "where she explains the pricing change"; "Find the part where the demo crashes"; "make a highlight reel" | Turn a webinar into short clips |
| Streams | Locating the plays in a multi-hour VOD | A description of the play; "make a highlight reel" | Stream VODs into gaming highlights |
| Courses and lectures | Finding the clean take and the chapter boundaries | "What's the main topic of this clip?"; a where-does-the-best-take-start question; "remove silences" | Remove silence without choppy cuts |
The podcast case is the clearest. An hour of conversation holds perhaps a dozen moments worth clipping, and the old way to find them is to listen to the hour. Search finds the moment you remember, a question confirms who said it, and the reel prompt proposes the peaks for you to accept or deny. For a backlog rather than one episode, Aura's batch editing runs across 50+ videos.
What are the limits?
Search is only as good as what you uploaded, answers are starting points, and the reel is a first cut.
- Search covers your library, nothing else. Aura searches the footage you uploaded to Loopdesk, not the web or anyone else's footage. If a shot was never recorded or never uploaded, no description will find it. Those uploads stay yours: Loopdesk never uses your videos to train AI models, content is encrypted in transit and at rest, and you can delete files at any time.
- Confirm quotes and names before publishing. A result or an answer tells you where to look, not what to publish. The demo answer counts two speakers; it does not name them. Before a clip goes out with a quote and a name on it, check the wording and attribution against the footage yourself.
- The reel is a starting point for the cut. "Highest-energy" is one scoring rule, and the quiet moment that matters most may not be a peak. Read the plan card, Deny it if the rule is wrong for the piece, and expect to adjust the timeline by hand.
- Found footage is real; generated footage is not evidence. You can fill a gap in a reel with B-roll from the Create panel (see Generate what you're missing). Treat that clip as illustration and label it as AI-generated. Our guide to C2PA Content Credentials for video explains how.
- Browser only. Loopdesk runs in the browser and needs an internet connection; there is no offline mode and no native app.
- AI operations use credits. AI operations draw on the credits included with your plan, with pay-as-you-go credits beyond the monthly allowance. Plans are on the pricing page; there is no free plan or trial.
Frequently Asked Questions
What is semantic video search in Loopdesk?
It is search by description rather than by file name or exact words. Type a moment the way you would say it, in any language, and Aura scans your auto-tagged library and returns the matching clips with exact timestamps. Click a result to add that moment to the timeline.
Do I need to tag or log my footage before I can search it?
No. Every upload is tagged automatically when it lands in Loopdesk, so a clip is searchable without manual logging, and a file you uploaded months ago is as findable as one you added today.
Can I search in a language other than English?
Yes. Describe the moment in any language and Aura searches for it. Captions are separate from search and are available in 108 languages, including 22 Indian languages.
What questions can I ask my footage?
What is covered, who says what and where the best take starts. In the demo, "What's the main topic of this clip?" is answered with "This clip discusses product-market fit for SaaS startups, featuring 2 speakers." Aura points you to the relevant moment with the answer.
How does "make a highlight reel" work?
You type the prompt on a take, Aura watches the footage and proposes a plan card, such as 4 peak moments taking 4:32 down to 0:42, with Accept and Deny buttons. Accept, and it cuts, joins and stitches the reel, ready to export. Deny, and nothing is cut.
Is a found moment or a highlight reel ready to publish?
Treat it as a starting point. Confirm quotes, names and numbers against the footage, adjust the timeline by hand where needed, then export at 1080p or 4K with no watermark. From there the Distribution Calendar schedules it and Performance Analytics tracks it.
Ready to stop scrubbing? Open Loopdesk, upload a long take and type the moment you are looking for.