video-transcripts
Get what is said in YouTube videos and Shorts, TikToks and Instagram reels as text, for one link or a creator's latest or most-viewed videos, then summarize it, pull hooks, claims and quotes, or repurpose it into posts and scripts. Use when the user shares a video link and asks what it says, wants a transcript or summary, or wants to study what a creator says across many videos.
- License 许可
- Apache-2.0
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- video-transcripts
- description
- Get what is said in YouTube videos and Shorts, TikToks and Instagram reels as text, for one link or a creator's latest or most-viewed videos, then summarize it, pull hooks, claims and quotes, or repurpose it into posts and scripts. Use when the user shares a video link and asks what it says, wants a transcript or summary, or wants to study what a creator says across many videos.
03 SKILL.md 正文
Video transcripts
Turn videos into text, then into something useful: a summary, the hooks and claims, quotes with timestamps, or a script to reuse. One link or a whole channel.
<!-- lurkapi:access -->
Calling LurkAPI
With MCP (the LurkAPI plugin or connector), call the tools by name, like tiktok_profile; your client may prefix them (mcp__plugin_lurkapi_lurkapi__tiktok_profile). With only find_endpoints and call_endpoint, run a tool as call_endpoint with endpoint set to its name and params to its arguments. Results come trimmed: no nulls, one URL per image or video. A long result comes in parts: its _note gives the offset for the next part, which is free.
Without MCP (Codex, Cursor, scripts), every tool is GET https://api.lurkapi.com plus its path, with the same params as a query string (URL-encoded) and the key from LURKAPI_API_KEY in the x-api-key header:
curl -s "https://api.lurkapi.com/v1/tiktok/profile?handle=nike" -H "x-api-key: $LURKAPI_API_KEY"
If LURKAPI_API_KEY isn't set but LURKAPI_KEY is (the name older docs used), use that. Raw answers are big: pipe them through jq to keep the fields you need. No key yet? Ask the user to create one at https://lurkapi.com/dashboard (100 free credits on sign-up, then a balance below 25 tops back up to 25 every day) and to set LURKAPI_API_KEY in their shell. Never ask for the key in the chat, and never print it.
Credits. Each call costs what the table at the end says, not_found included; failures on our side (5xx) are free. Every answer carries credits_remaining. Before a run of more than 20 calls, check the balance with account_credit_balance (free) and tell the user what the run will cost. If a call answers missing_api_key, insufficient_credits or signup_required, stop and pass on what it says.
<!-- /lurkapi:access -->
One video
Pick the call by the link:
- YouTube (watch, youtu.be, /shorts/, /live/):
youtube_transcript(url=<url>). The text istranscript_only_text; the timed version istranscript(each part withstartTimeText), which MCP returns instead withtimestamps=true.language=espicks a caption track fromavailableLanguages.isGeneratedtrue means YouTube's auto-captions: expect misheard names. - TikTok:
tiktok_transcript(url=<url>). The text istext. Atranscriptof null with emptylanguagesmeans the video has no captions;use_ai_as_fallback=truethen transcribes the audio itself (videos up to 2 minutes), which costs more (see the table). Ask before turning it on for many videos. When you also want the video's views and likes,tiktok_video(url=<url>, get_transcript=true)gives stats and captions in one call, as WEBVTT. - Instagram reels:
instagram_media_transcript(url=<url>). Reels have no captions to read, so a new one always goes through speech-to-text at the higher price in the table; one we've transcribed before costs the base price. Up to 120 seconds. - Anything else (Facebook, X, podcasts): not supported. Say so.
A video with no speech (music only) answers transcript_unavailable, at the base price.
Many videos
List first, then transcribe only the videos that matter.
- YouTube:
youtube_channel_videos(handle=<handle>, sort=latest)(about 30 a page;sort=popularfor the most viewed), oryoutube_channel_shorts(handle=<handle>). Thenyoutube_transcripton the chosenurls. - TikTok:
tiktok_profile_videos(handle=<handle>): 15 a page, newest first, withstats.playCount. Thentiktok_transcripton the chosen ones. - Instagram:
instagram_user_reels(username=<username>)gives play counts, theninstagram_media_transcript(code=<code>)per reel. At the speech-to-text price, ten reels cost more than ten captioned TikToks: say the total first.
To find a creator's best videos rather than their latest, rank the list by views and take the top few, or the ones far above the creator's median.
Budget
A transcript is one call; a listing page is one more. "The last 10 TikToks" is 11 calls; "the top 5 videos on a channel" is 6. Instagram reels and TikTok's audio fallback cost more per video: work it out from the table and say it before a batch.
What to make of it
Unless the user asks for something else, give:
- Summary: what the video argues or shows, in three sentences.
- Hook: the first two sentences, verbatim, and why they work.
- Structure: the beats in order, with timestamps when you have them.
- Claims: each factual claim or number, quoted, so it can be checked.
- Call to action: what the viewer is asked to do.
- Reusable lines: quotes that stand alone.
Across many videos, compare hooks, recurring topics and phrases, video length against views, and how the best performers open.
Pitfalls
- Never quote what isn't in the transcript. Auto-captions mishear names, brands and numbers; flag doubtful ones.
- A transcript is what's said, not what's shown: on-screen text and visuals aren't in it.
- Asking for the same transcript again is charged again, at the base price: keep what you fetched.
<!-- lurkapi:calls -->
Calls and credits
| Tool | REST path | Credits |
|---|---|---|
youtube_transcript |
/v1/youtube/video/transcript |
1 credit |
tiktok_transcript |
/v1/tiktok/video/transcript |
1 credit; 2 if speech-to-text runs |
tiktok_video |
/v2/tiktok/video |
1 credit |
instagram_media_transcript |
/v2/instagram/media/transcript |
1 credit; 2 if speech-to-text runs |
youtube_channel_videos |
/v1/youtube/channel-videos |
1 credit |
youtube_channel_shorts |
/v1/youtube/channel/shorts |
1 credit |
tiktok_profile_videos |
/v3/tiktok/profile/videos |
1 credit |
instagram_user_reels |
/v1/instagram/user/reels |
1 credit |
Each tool's link opens its docs as markdown: every param and its values, and a real response. <!-- /lurkapi:calls -->
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.