Visitar URL original
Search - Arcmira API
Skip to main content
Search what people say across indexed YouTube videos and podcasts. Ask your agent:
  • “Find quotes about GPU financing in the last 60 days, with speakers and source links.”
  • “Find Scott Wu’s media appearances across YouTube and podcasts.”
  • “Which companies are gaining attention, and what are people saying about them?”
Search transcript text with a word or phrase. To filter by a person, company, product, topic, or channel, resolve its name first and pass the returned ID.

Search YouTube videos and podcast transcripts for quotes and clips

GET /v1/search returns passage text, source metadata, publication dates, and timestamped watch links. This example searches the latest 60 days:
Selected fields from the first passage returned in an October 17, 2026 capture. The passage text is complete; the date window uses UTC. Your results will vary.
This sample omits other response fields and the remaining passages. Keep hits.chunks for further analysis. Open a passage’s watch_url or read the surrounding transcript before drawing a conclusion. Follow the query into investment research with YouTube video and podcast transcript search. An empty search means no matching passages in the indexed coverage and returned window. If a source you need is missing, help shape the index. Include source links and the workflow better coverage would support.

Find mentions of an entity

Find where people discuss a person, company, product, or topic. Mention records describe where a subject came up. Use transcript search when you need the spoken words. Resolve PlanetScale, then request its mentions with GET /v1/mentions?entity_id=<resolved-id>. Apply the same after and before window used for your research.

Resolve a name to an ID

Filters take ids, never names. GET /v1/entities/resolve turns a name, @handle, YouTube URL or UC... id into one entity. It is free. The answer is one of three: None of the three means the name is not indexed. Pass context with the user’s own words (“the startup bank”) to settle close calls. Pass type=channel for a show and use its youtube_channel_id; omit type for a brand. Mention rows include a description, source video, and available timestamps. Read the description before asserting a match because names can refer to different subjects. start_seconds and end_seconds describe the mention; a whole-episode record can use 0.

Find a person’s media appearances

Arcmira records appearances as a type of mention. Resolve Scott Wu with type=person, then query GET /v1/mentions with his entity ID and the date window. Each result is an indexed video. Separate clips from an interview can appear as separate results. Search transcript passages mentioning a subject for specific words or phrases. Use about with the entity ID, by for a person’s own words, and channel or date filters to narrow the evidence. A name in an ID filter answers 400 id_required. Read the returned window to retain the dates the search applied.

Check mention momentum

See whether an entity is gaining attention and where to investigate next. GET /v1/entities/{id}/momentum?days=7 compares the latest seven days available on your plan with the preceding seven. Choose 7 to 365 days; the default is 30. The response includes current and prior mention counts, absolute and percentage change, a momentum verdict, the latest source date, and top shows. Search the entity’s transcripts to learn whether increased attention reflects enthusiasm, criticism, sponsorships, or something else. See which companies are gaining attention in Famous Rankings. Explore rising companies, the shows discussing them, and eight weeks of history to find your next research lead. Momentum measures mention volume. Famous Rankings uses its own ranking method. Each momentum lookup uses credits from your plan; see Plans and usage.

Count mentions and appearances across shows

GET /v1/mentions/counts counts entities across selected channels, entities, or videos. Set a date window; counts are all-time without one. Set mode to mentions, appearances, or both, and limit from 1 to 40. Set entity_types=topic for subjects, person for people, or organization,product for brands. Each row’s count is the number of episodes in which the entity appears; occurrences counts the catalog occurrences. With two or more channels, shared lists the entities they have in common. To find a show’s latest videos, use GET /v1/channels/{channel_id}/videos. Passages and list rows use credits from your plan past the first five of a request; see Plans and usage.

Time windows

Agents default to the last 90 days when you name no window. Over HTTP, lists run newest first with no window unless you set one. Every dated route takes after and before. The window is half-open: after is included and before is not, so September is after=2026-09-01&before=2026-10-01. Each takes a date or an ISO 8601 datetime with an offset, read in UTC. The body echoes the window it applied as window. On GET /v1/mentions, before needs a paid plan. The Free plan withholds the newest 30 days. as_of on a search is the newest returned passage’s date, not how fresh the index is. For that, read search_indexed_through from GET /v1/channels/{channel_id}/coverage. Spot a wrong row or need more coverage? Help shape Arcmira. Include the source or entity ID, your workflow, and what would make the results more useful.