The Discovery Agent is built to understand natural language—so you don’t need to use strict keywords or know how your archive is organized. But like any AI assistant, how you ask can shape what you get.

This guide shares tips to help you write better prompts, understand where the Discovery Agent excels, and work around its current limitations to get straight to the moments you're looking for.

Understanding What the Discovery Agent Does Best

The Discovery Agent works by analyzing visual content, contextual cues and spoken dialogue. Knowing where it excels—and where it currently has limitations—helps you craft prompts that deliver the best results.

✅ Where Discovery Agent Works Especially Well

Searching for people doing or saying things	The Discovery Agent excels at finding specific people or recurring subjects, especially in well-catalogued content.	"The President announcing new policies at press conferences" "Messi celebrating goals" "Monica being overly competitive in Friends" "The CEO talking about company values in town halls"
Visual settings and environments	If you can describe the location or setting visually, the Agent can find it.	"Protest scenes in city centers" "Packed stadium atmosphere shots during evening matches" "Romantic dinner scenes at sunset" "Product being used in outdoor environments"
Clear visual actions	The Agent is strong at identifying moments based on what's happening on screen.	"Emergency responders arriving at incident scenes" "Penalty kick sequences" "Dance sequences in musicals" "Fight scenes in action movies" "Assembly line workers packaging products"
Thematic and mood-based searches	The Agent can recognize broader themes using visual and contextual storytelling elements.	"Emotional reunion moments," "Tense debate exchanges" "Dramatic last-minute goals," "Injured players leaving the field" "Dramatic courtroom moments from legal dramas" "Team collaboration moments in office settings"

Note: The Discovery Agent works especially well on widely recognized shows, sports leagues and major news events. The AI reasoning model has learned from iconic content and memorable moments across industries.

⚠️ Where Discovery Agent Has Limitations

Cinematography techniques	The Discovery Agent cannot currently detect specific camera techniques or shooting styles.	"Slow motion replays" "Shots filmed from low angles" "Drone footage with tracking shots"
Highly specific movements or techniques	General movements work well, but nuanced or technical body language is less reliable.	"A Cruyff turn in football," "Specific boxing combinations" "Precise hand gestures during testimony"
Dialogue tone and delivery	The Discovery Agent cannot currently detect sarcasm, irony, politeness or vocal tone.	"Politician responding sarcastically to questions" "Commentator making ironic remarks" "Polite customer complaints"
Subjective or abstract qualities without context	Humor, intensity and similar subjective traits are highly personal and require training data or examples to work well.	"Funniest moments" "Most intense plays" "Awkward interview moments"

Start with the why

Good prompts don’t just describe what you want—they explain why you want it. When the Agent understands the purpose, it delivers more relevant moments.

Here are some good examples of prompts that provide great context by explaining the type of content you are looking to create.

“I’m making a best-of compilation, so I need short, memorable quotes that sum up key achievements.”

“I’m cutting a trailer, and need dynamic crowd reactions to match a fast-paced edit.”

“I’m a researcher for a documentary about climate change, looking for insightful interviews and key visuals that explain the positive impact of wind turbines on reducing greenhouse gases.”

“I’m putting together an event recap, so I’m looking for speaker highlights, applause moments, and people networking.”

Tip: Add your use case—it helps the Agent prioritize tone, pace, and framing.

Give it context

The more background you provide, the better the Discovery Agent can tailor its search and deliver exactly what you need.

Think about who your audience is, where the content will be published, and the style or format you usually work with. Sharing these details helps the agent understand the bigger picture.

For example, if you’re creating content for:

Social media platforms like Instagram or TikTok where short, punchy clips work best
A documentary for fast channels requiring in-depth interviews and informative visuals

Include those details in your prompt.

If you have a creative brief or a set of instructions from your team, copy that into your query too. This helps the agent grasp your intent and focus on what matters most.

Be natural, but precise

The Discovery Agent is smart—but it’s not a mind reader. Vague or overly subjective prompts like “something cool” or “a nice clip” are open to interpretation and could lead to underwhelming results.

Here are some good examples of prompts that better guidance to the Discovery Agent by detailing types of moments or clips you are looking for rather than possible subjective statements.

“Looking for quotes where someone talks about overcoming self-doubt or imposter syndrome—ideally in a reflective tone.”

“Clips showing crowded crosswalks, street vendors, traffic during rush hour—no skyline shots.”

“Looking for moments where people are laughing, dancing, or celebrating in a casual setting.”

“Find moments where the audience erupts in applause, or the keynote speaker delivers a bold, memorable line.”

Tip: Break down abstract ideas into visuals, soundbites, or feelings.

Tell it what kind of content you want

Be clear about the format or type of moment you’re after. The Discovery Agent can return a sequence, a shot, a soundbite or even full videos—but it helps to be specific.

Looking for…	Try saying…
A single shot (e.g. visual b-roll)	“I need a wide shot of a crowded train platform during rush hour—no dialogue, just ambient.”
A moment (a short stretch of narrative or action)	“Show moments of someone prepping food in a kitchen—from chopping to plating.”
A photo (a still image)	“A photo of a busy urban intersection at night—long exposure with light trails from traffic.”
A soundbite (short spoken quote or moment)	“Find a soundbite where someone talks about overcoming self-doubt—under 10 seconds, emotionally delivered.”
A full video (self-contained clip or story)	“Do we have a full video on our latest community outreach project—ideally under 2 minutes?”

Use follow-ups to narrow your search

Think of it as a conversation. You don’t need to restart from scratch—just refine.

Include the following instructions straight up in your prompt or use them as follow-ups:

“Only show clips that are 3 minutes in duration”
“Show aerial shots, wide shots, close up shots”
“Only show results from the 1990s, only show results from last week, month or year”

Tip: The Agent remembers your previous prompt—so you can iterate naturally.

If you have any further questions, don’t hesitate to reach out to us at support@momentslab.com — our team will be happy to help.

How to Prompt the Discovery Agent