# VidWords > VidWords (https://vidwords.com) turns YouTube videos, channels and playlists into transcripts and grounded video research. Browser transcripts are unlimited and free; work performed by VidWords uses separate Cloud Request and AI Processing Unit balances. A free account includes 25 monthly Cloud Requests, 200 monthly AI Units and 100 one-time Developer Requests. There are also AI summaries and chat, a developer REST API and MCP server, Radar monitoring, and VidWords Watch — an AI video journey combining chapter summaries, real frames, transcript excerpts, study notes and evidence-linked Q&A. ## Core pages - [YouTube Transcript Extractor](https://vidwords.com/): paste a YouTube link, get the transcript - [AI Watch — AI video summarizer](https://vidwords.com/ai-watch): paste a link, get a chapter-by-chapter AI journey with real frames, transcript excerpts, study notes and citation-checked Q&A. Standard Watch uses 3 AI Processing Units per minute of video; Deep Watch uses 30 and is included on Pro and Team. Free includes 200 AI Units a month. - [Radar — YouTube channel watchlists](https://vidwords.com/radar): monitor chosen channels for new uploads. Polling itself is free; retrieving a matched transcript uses one Cloud Request and optional AI analysis uses AI Processing Units. Free supports 1 daily watchlist, Starter 3 daily, Pro 25 roughly every four hours, and Team 100 hourly. - [Bulk extraction](https://vidwords.com/bulk): playlists, channels, and lists of links at once - [API documentation](https://vidwords.com/api-docs): POST /api/transcripts (up to 50 videos per call) and POST /api/channels; Basic auth with a free API token - [OpenAPI 3.0 specification](https://vidwords.com/openapi.json): the machine-readable spec — every endpoint, request and response field, and error code, ready to import into a client generator or drive directly from an agent - [Video Evidence API](https://vidwords.com/api-docs#watch): POST https://vidwords.com/api/watch starts frame-level analysis, GET /api/watch/:jobId polls it, and POST /api/watch/:jobId/ask asks an evidence-grounded question. Standard analysis uses 3 AI Processing Units per minute of video; Deep uses 30 and is included on Pro and Team. Refused jobs cost nothing and failed charged jobs are refunded. - [MCP server](https://vidwords.com/api-docs#mcp): a hosted Model Context Protocol endpoint at POST https://vidwords.com/mcp. It authenticates with the same API token as REST and uses the same entitlements: transcript retrieval costs one Cloud Request per video; Watch and grounded questions use AI Processing Units. Every new account gets 100 Developer Requests, and PAYG Cloud packs do not require a subscription. - [Pricing](https://vidwords.com/pricing): Free, Starter ($10/mo or $100/yr), Pro ($29/mo or $290/yr), and Team ($79/mo or $790/yr). Monthly Cloud Requests: Free 25, Starter 1,000, Pro 3,500, Team 10,000. Monthly AI Processing Units: Free 200, Starter 3,000, Pro 10,000, Team 30,000. Browser transcripts are unlimited. Non-expiring Cloud and AI top-ups are available on Free. - [Free YouTube Transcript Tools](https://vidwords.com/tools): three browser tools that run entirely client-side on transcript text you already have — a cleaner, a word/reading-time counter and an AI prompt generator. Format EXPORTS are not here: they live in the extractor itself (.txt/.srt/.vtt/.csv/.json from the download menu) and are documented in the /resources guides - [YouTube Transcript Cleaner](https://vidwords.com/tools/clean-transcript): strip timestamps, remove sound tags like [Music], fix sentence casing, and merge short lines into readable paragraphs - [YouTube Word Counter & Reading Time](https://vidwords.com/tools/word-counter): calculate spoken words, words per minute, silent reading duration, and time saved by reading vs watching - [AI Video Prompt Generator](https://vidwords.com/tools/prompt-generator): create tailored prompts for ChatGPT, Claude, and Gemini to generate study notes, Anki flashcards, blog posts, and quizzes - [Chrome extension](https://vidwords.com/extension): reads the transcript in a side panel next to the video, with search, caption-language switching, translation, TXT/SRT/VTT export and timestamped notes that save to the account. Published on the Chrome Web Store — install it at https://chromewebstore.google.com/detail/njjmjcjjcplanbdliidknpaepifknohh . A hand-installable ZIP is also served for browsers that cannot reach the store, but it does not auto-update - [AI Watch demo — a real, unedited run](https://vidwords.com/watch-demo): the actual output of one AI Watch analysis of Apple's public 76-minute WWDC 2026 keynote — chapter by chapter, with real video frames and clickable timestamps. Nothing on the page was hand-written or edited, so it is the page to check when the citation and frame-reading claims above need verifying rather than taking on trust. - [About VidWords](https://vidwords.com/about): how the product works, what it stores, known limits, and how its guides are maintained ## Guides - [How to extract a YouTube transcript](https://vidwords.com/resources/how-to-extract-youtube-transcripts): Extract available captions from a public YouTube video with timestamps, chapters, language switching, and TXT/SRT/VTT/CSV/JSON export. - [Convert YouTube video to text](https://vidwords.com/resources/youtube-video-to-text): Convert any YouTube video to text in seconds. Paste a URL, get clean readable paragraphs with optional timestamps and chapters, then copy or download as TXT. - [Download YouTube subtitles as SRT/VTT](https://vidwords.com/resources/download-youtube-subtitles-srt): Export available YouTube captions as SRT, VTT, TXT, CSV, or JSON. Choose a caption language, preserve timestamps, or process a playlist in bulk. - [Summarize YouTube videos with AI](https://vidwords.com/resources/summarize-youtube-videos-ai): How to summarize any YouTube video with AI: get the transcript, generate key points in one click, or paste it into Claude/ChatGPT with ready-made prompts. - [VidWords vs Eightify vs NoteGPT](https://vidwords.com/resources/vidwords-vs-eightify-notegpt): An honest Eightify alternative comparison: VidWords Watch vs Eightify vs NoteGPT on video frames, verbatim quotes, citations, free tiers, and price. - [Make flashcards from a YouTube video](https://vidwords.com/resources/youtube-video-flashcards): Turn any YouTube video into flashcards: generate a self-test with AI in one click, then export Markdown for Anki or Notion. - [YouTube transcript to ChatGPT](https://vidwords.com/resources/youtube-transcript-to-chatgpt): Copy a YouTube transcript into ChatGPT or Claude to summarize, ask questions, or repurpose it. Step-by-step workflow, ready-made prompts, and a long-video fix. - [YouTube transcript API for developers](https://vidwords.com/resources/youtube-transcript-api-developers): Get YouTube transcripts programmatically: why the Data API won't return caption text, open-source vs hosted options, plus Python and JavaScript examples. - [YouTube auto-captions explained](https://vidwords.com/resources/youtube-auto-captions-explained): How YouTube auto captions actually work, how accurate they really are, why some videos have no transcript at all, and what you can do in each case. - [Turn YouTube videos into blog posts](https://vidwords.com/resources/turn-youtube-videos-into-blog-posts): A practical workflow to turn a YouTube video into a blog post: extract the transcript, clean it up, restructure it with an LLM, edit for reading, and publish. - [YouTube transcripts for students](https://vidwords.com/resources/youtube-transcripts-for-students): Turn YouTube lectures and course videos into searchable notes: skim hours in minutes, quote with timestamps, build study guides, and cite sources properly. - [YouTube transcripts for research](https://vidwords.com/resources/youtube-transcripts-for-research): A guide for researchers and journalists: build a corpus from YouTube videos, export CSV/JSON for NVivo or Atlas.ti, and cite exact timestamped moments. - [Best free YouTube transcript tools compared](https://vidwords.com/resources/best-free-transcript-tools): Every free route to a YouTube transcript — YouTube's own panel, open-source libraries, why the official Data API cannot return caption text — plus a checked side-by-side of the hosted tools (Tactiq, NoteGPT, TurboScribe, Descript, DownSub, Kome, VidWords) on free tier, export formats, batch size and API access. Leads with the distinction the category hides: caption extractors read a file YouTube already stores and are helpless without one, speech-recognition services re-transcribe the audio and therefore price and cap by duration. - [A Tactiq alternative for YouTube transcripts](https://vidwords.com/resources/tactiq-alternative): Tactiq is a live meeting transcriber (Meet, Zoom, Teams) whose free YouTube generator is a standalone side tool — no signup, TXT only, one video at a time, no documented API. Says plainly that Tactiq is the better purchase when the real problem is meetings, and compares the YouTube-specific surface: exports, playlists and channels, search, and MCP access. - [A TurboScribe alternative for YouTube videos](https://vidwords.com/resources/turboscribe-alternative): TurboScribe runs speech recognition over audio; VidWords reads the existing caption track. Covers the free plan's 3-files-per-day and ~30-minutes-per-file ceiling (which makes a two-hour lecture impossible on the free tier), where TurboScribe genuinely wins — captionless videos, bad automatic captions, non-YouTube media, speaker labels — and where caption extraction is the cheaper architecture. Notes that turboscribe.ai returns 403 to automated requests, so its figures are third-party-sourced and should be verified. - [Descript for YouTube transcripts, and when it is the wrong tool](https://vidwords.com/resources/descript-alternative): Descript is a text-based video and audio editor that transcribes as a side effect; its own llms.txt never mentions YouTube. Compares the media-minute free tier (60 minutes a month, total) and the $24/mo entry price against a transcript extractor's, and gives the deciding question: are you going to change the video? - [A NoteGPT alternative for YouTube transcripts](https://vidwords.com/resources/notegpt-alternative): The closest genuine competitor — a broad AI study suite (flashcards, quizzes, slides, podcasts, PDFs) versus a narrow transcript and evidence tool. Compares exports (TXT vs five formats), batch size (20 videos vs 500), published free quotas versus unstated ones, developer surface, and citation verification. - [Bulk transcripts for a whole channel or playlist](https://vidwords.com/resources/channel-transcripts): Extract transcripts for an entire YouTube channel or playlist from one handle or link — in the browser or over the API, up to 500 videos at a time. - [Search a whole channel's transcripts for a topic](https://vidwords.com/resources/search-youtube-channel-transcripts): Find where a topic is mentioned across every video in a channel and get back the quote plus a clickable youtube.com/watch?v=…&t=…s link, rather than a pile of transcript files. Covers browser, REST and MCP routes, the literal-vs-semantic search choice, and Cloud Request usage. - [Analyse what is on screen in a YouTube video](https://vidwords.com/resources/analyze-youtube-video-on-screen): Read slides, charts, code, UI state and on-screen text — the things a caption track never contains. Standard Watch uses 3 AI Processing Units per analysed minute and verifies citations against stored evidence. - [Bulk YouTube transcript downloader](https://vidwords.com/resources/bulk-youtube-transcript-downloader): The four ways to hand over a list of videos (playlist, channel handle, pasted links, CSV upload), the real per-request caps, what the combined TXT/CSV/JSON exports each contain, and an honest comparison against looping yt-dlp yourself. - [YouTube transcript not showing](https://vidwords.com/resources/youtube-transcript-not-showing): Symptom-first troubleshooting for a missing, greyed-out, blank or wrong-language transcript button — six distinct causes with the fix for each, plus what to do when the video genuinely has no caption track and audio transcription is the only route. - [YouTube transcript to Word (.docx)](https://vidwords.com/resources/youtube-transcript-to-word): Getting a transcript into Microsoft Word — which view to copy from to get prose rather than timestamps, how to convert the TXT export without wrecking the encoding, and what the one-click Watch .docx export actually contains. - [Fixing RequestBlocked and IpBlocked in youtube-transcript-api](https://vidwords.com/resources/youtube-transcript-api-ip-blocked): Why the open-source library works locally and fails on any cloud host, the fixes that do and do not change the outcome (including why a pooled requests.Session stops a rotating proxy from rotating), the PoTokenRequired wall no proxy solves, and when a hosted API is the right call versus overkill. - [YouTube transcripts on AWS Lambda](https://vidwords.com/resources/youtube-transcript-api-aws-lambda): Why Lambda fails intermittently rather than consistently, why the NAT-gateway-plus-Elastic-IP fix converts a partial failure into a total one, and the 3-second timeout and 250MB package limit that get misreported as IP bans. - [YouTube transcripts on Render, Railway and Heroku](https://vidwords.com/resources/youtube-transcript-api-render): On a long-running host the outbound IP is static and shared per region, so the block is permanent and redeploying cannot clear it — why dedicated outbound IPs make it worse, and the caching and rate-limiting a persistent process can do that serverless cannot. - [youtube-transcript-api in Google Colab](https://vidwords.com/resources/youtube-transcript-api-google-colab): Why notebooks are blocked more often than servers, why restarting the runtime is a coin flip rather than a fix, keeping proxy credentials out of a shared .ipynb with Colab Secrets, and caching to Drive so a deleted VM does not mean refetching everything. - [n8n / Make.com / API integrations](https://vidwords.com/resources/integrations): Connect YouTube transcripts to your automations: ready-made n8n and Make.com templates, plus REST API recipes for custom pipelines. - [A free youtube-transcript.io alternative](https://vidwords.com/resources/youtube-transcript-io-alternative): A free youtube-transcript.io alternative: VidWords gives 25 free transcripts a month with a free account (3 with no signup), 5 export formats, bulk playlists, AI summaries, and an API. - [A youtubetotranscript.com alternative](https://vidwords.com/resources/youtubetotranscript-alternative): A youtubetotranscript.com alternative comparing 5 export formats (TXT/SRT/VTT/CSV/JSON), automatic speech-to-text rescue for missing captions, AI Watch frame analysis, bulk playlist extraction, and an ad-free interface. - [Why ChatGPT can't watch a YouTube video](https://vidwords.com/resources/chatgpt-cant-watch-youtube-videos): Paste a YouTube link into ChatGPT and it usually can't watch the video. What actually happens, why a URL alone isn't enough, and the transcript workflow that fixes it. - [Best YouTube transcript APIs compared](https://vidwords.com/resources/best-youtube-transcript-api): An honest comparison of YouTube transcript APIs in 2026 — pricing, batch limits, captionless-video fallbacks and rate limits, with the criteria that actually matter. - [YouTube MCP server for AI agents](https://vidwords.com/resources/youtube-mcp-server): the hosted Model Context Protocol endpoint at POST https://vidwords.com/mcp explained end to end — why a grounded, citable answer rather than a raw transcript is the useful unit to hand an agent, all nine tools with their costs, Basic-token auth and the email-verification gotcha, why search_transcript is preferred over get_transcript, the 30-per-10s rate limit and why it is looser than REST, frame-level analysis via analyze_video/ask_video, a drop-in agent skill at github.com/haljishi/vidwords-mcp/blob/main/skills/youtube-transcripts/SKILL.md carrying the tool-selection guidance (prefer search over get, pass from/to timecodes, send a list of ids for a channel), and the honest limits (header auth only with no OAuth flow, stateless, RapidAPI identities refused). - [YouTube MCP for Claude Code](https://vidwords.com/resources/youtube-mcp-claude-code): connect Claude Code with one `claude mcp add --transport http` command, choose between user and project scope, and avoid committing an API token in a project-scoped .mcp.json. Includes what to ask for and why screen-recorded tutorials need frame reading rather than captions. - [Connect YouTube to Claude Desktop](https://vidwords.com/resources/youtube-mcp-claude): the claude_desktop_config.json block with file paths for macOS and Windows, why a full quit rather than closing the window is required, how to merge with existing mcpServers entries without invalidating the JSON, and troubleshooting. - [YouTube MCP for Cursor](https://vidwords.com/resources/youtube-mcp-cursor): the .cursor/mcp.json block, project versus global config, keeping the API token out of version control, and using the tools from Agent mode. - [How to get a YouTube Shorts transcript](https://vidwords.com/resources/youtube-shorts-transcript): Extract captions from vertical Shorts — mobile app menu, desktop URL conversion (/shorts/ to /watch?v=), and TXT/SRT export. - [YouTube transcript to Notion](https://vidwords.com/resources/youtube-transcript-to-notion): Clean markdown notes, chapter toggle lists, study templates, and automated database sync via n8n and Make. - [YouTube transcripts in Python](https://vidwords.com/resources/youtube-transcript-python): Python guide comparing youtube-transcript-api vs VidWords REST API, handling IpBlocked/bot-wall errors in cloud environments, and parsing JSON/SRT. - [Copy YouTube transcript without timestamps](https://vidwords.com/resources/youtube-transcript-without-timestamps): Three ways to remove timecodes for clean reading, word documents, and token-efficient AI prompts — including which VidWords layout is the timestamp-free one. - [Fix TranscriptsDisabled and NoTranscriptFound in Python](https://vidwords.com/resources/youtube-transcript-api-transcripts-disabled): Why youtube-transcript-api raises TranscriptsDisabled, NoTranscriptFound, or CouldNotRetrieveTranscript, and how to safely catch and fall back in Python. - [Fix yt-dlp bot verification](https://vidwords.com/resources/yt-dlp-bot-verification): Why yt-dlp fails with Sign in to confirm you're not a bot or BotVerificationRequired on cloud servers, and the proven workarounds that work in production. - [YouTube transcripts in LangChain](https://vidwords.com/resources/youtube-transcript-langchain): How to load YouTube transcripts into LangChain for RAG pipelines, timestamp-preserving chunking, and Chroma vector search. - [YouTube transcripts in LlamaIndex](https://vidwords.com/resources/youtube-transcript-llamaindex): Build video indexing and Q&A retrieval engines with LlamaIndex and timestamped node metadata. - [All guides](https://vidwords.com/resources) ## API quick reference POST https://vidwords.com/api/transcripts with JSON body {"ids": ["