Notetaker#

Turn videos, podcasts and recordings into accurate transcripts. Paste a YouTube or Bilibili video link, a direct link to a podcast episode or audio file, or upload an audio or video file, and Waper writes the transcript, a summary and chapters, translated into your language. Every passage has a timestamp you can play back from; save it to a knowledge base, and your agent can keep working with it.

What it does and which platforms it supports: Video to text on waper.ai.

Where to find it#

• Sidebar: Notetaker sits between Radar and Library. On a phone it's in the bottom bar, whose five tabs are Chats, Radar, Notetaker, Knowledge base and Me.
• Search: type "Notetaker" in ⌘K to go there. Titles of your audio and video and lines in their transcripts are searchable too; a transcript match opens at that line.
• Chat: send the link to an agent and ask it to transcribe; see "Transcribing from a chat" below.
• iPhone app: Notetaker in the tab bar, where you can paste a link to transcribe and see all your audio and video; upload files from your computer on the web. See iPhone app.

How to start#

1. On the Notetaker home, paste a link into the box at the top (a whole paragraph works; the first link in it is used). Roughly how many minutes it runs shows right away, and your first recording also says its first 60 minutes are free. Or drop a file on the page, or select "Upload" to choose one (one at a time, up to 2 GB);
2. Nothing starts until you select "Start": nothing is charged and nothing is processed before that. The page opens on the item, with the progress at the top;
3. If someone already transcribed the link, it says a transcript is already there; it is reused and no credits are used.

Supported: YouTube, Bilibili, direct links to podcast episodes and audio files, and audio or video files from your computer. Playlists, live streams and paid content aren't supported.

Below the box is all the audio and video in your library, grouped by date, with a filter by source (All, YouTube, Bilibili, Podcasts & audio, Uploads). Each row shows the source, length and knowledge base; one in progress shows its current step, and a failed one says why and can be retried.

When processing finishes, existing feed episode titles or upload filenames are kept. Direct audio links without a title use the filename without its extension.

Progress#

The page shows three steps: getting the content → turning it into text → writing the summary. The transcript appears as it goes, so you don't need to wait around: it'll show up in Notetaker (and your library) when it's ready, and you'll get a reminder from Kiki in its chat.

If the service restarts while preparing the audio, the job prepares it again automatically. The upload is kept until preparation completes so processing can continue, within the temporary upload's 12-hour retention period.

While processing, the length may be an estimate. Once audio is decoded, its actual length is used consistently in the completed item’s details, Notetaker list and Library list.

The page for one recording#

On a wide screen the player, title, source and chapters are on the left, and the "Summary" and "Transcript" tabs on the right. On a phone the player is at the top and shrinks into a mini player bar as you scroll.

Play back#

Select any timestamp (a key point, a chapter, a transcript passage) and the player jumps there:

• YouTube and direct audio links: it jumps and plays, and the passage being played is highlighted. "Follow" is on by default, so the transcript scrolls with playback; scrolling it yourself pauses following, and "Back to what's playing" picks it up again;
• Bilibili: the player reloads at that second, without highlighting or following;
• Uploads: the original file isn't kept after processing, so there's no player; timestamps move you around the transcript.

Summary#

• Overview: what the piece says in one paragraph;
• Key points: bulleted, each with a timestamp;
• Chapters: thematic sections, each with a title, a time and one line.

Below are a few follow-ups (pull out action items, write it up as an article in a knowledge base, list notable quotes); selecting one hands it to your agent.

Transcript#

• Searchable, with a timestamp on every passage;
• Typos and meaningless filler are cleaned up without changing order or meaning; for a corrected passage you can "View machine original";
• When the transcript isn't in your language, switch between Original, Translation and Bilingual (each line with its translation below);
• Select one or more passages to "Add to note" or "Copy with timestamps".

For direct audio links and uploads, Waper prefers the file's title, show name and author tags. Without tags, it uses the filename without the link's query parameters. The price may initially show a filename; tags are filled in after downloading.

Save to a knowledge base#

Each recording gets one note of your own:

• The first time you select "Save to knowledge base", Waper creates a note document (source, overview, key points and chapters, with timestamps that link to that moment on the original video or podcast; an uploaded file has no original page, so its timestamps are plain text) in the item's knowledge base; after that the button reads "Open note";
• Select lines in the transcript and "Add to note" to append them to the end of the same note as quotes with timestamps (linked to the original the same way); if there's no note yet, one is created first;
• While the summary is still being written, both buttons wait and turn on once it's done; if the summary couldn't be written, you can still save, and the note just leaves out the overview, key points and chapters;
• The note is an ordinary document you can edit freely.

Ask your agent#

The "Ask" button at the top right names who you'll ask: opened beside a chat, it's that chat's agent and the cursor goes straight into the chat's input box; opened on its own, it's the agent that transcribed it (your main agent if it wasn't done in a chat), opened beside the page. Either way it has this recording as context, so you can ask about it; citations carry timestamp links like [12:30]. The agent can go on to save the result to a knowledge base or set up a watch.

More#

"More" has Open original, Export (Markdown, SRT subtitles, TXT), Move to knowledge base and Delete. Markdown includes the overview, key points, chapters and the timestamped transcript; SRT uses each passage's start and end times; TXT is one passage per line with its timestamp.

Translation#

When the transcript or summary isn't in your content language, the page shows it translated into your content language by default (passage by passage, same timestamps), and you can switch back to the original or to bilingual any time. Each language is translated once and shared, at no charge.

Pricing#

One price by length: 6 credits a minute (each started minute counts). The transcript, the summary and the translation into your language are all included.

• Your first recording is free: the first one you start on an account is on Waper for up to 60 minutes, even with a balance of 0; anything past 60 minutes is charged by the minute as usual. The price says so. It needs a verified email, and if the first one doesn't go through it isn't used up (see "Getting started" in Plans & credits);
• You always see the length before it starts, and it starts only when you confirm; for the per-minute charge see Plans and credits;
• You're only charged when it succeeds; uploads are settled on their real length. If starting fails, reserved credits are released. Retrying the same upload request, including overlapping retries, returns its existing record without starting another task or requiring additional available credits;
• Each link is processed once: if someone already did it, you get it free — and a summary or translation in your language is added at no charge.

Transcribing from a chat#

When you ask your agent to transcribe a link, the chat shows a progress card. When it would cost more than the credits you have left, or a lot at once, the agent shows a price card first and starts only when you confirm. Once the transcript is ready, the agent continues in the same chat with what you asked for (a summary, notes, an answer); you don't need to ask again. The recording shows up in Notetaker and your library as well.

Audio and video on the radar#

Videos and podcasts picked into a brief get transcripts filled in behind the scenes (not charged to you), and the entry's summary is rewritten from the transcript. Open it with "Transcript & chapters" on the brief entry; it appears in Notetaker once you save it.

Agents, MCP and the API#

Your agent, MCP and the API can all read transcripts and start a transcription (charged the same way). See MCP and API.

Limits#

Free handles recordings up to 60 minutes each; Personal and Pro go by credits rather than a plan cap, up to 10 hours each (which only keeps out unusually large files). Uploads included. Background transcriptions from your watches aren't held to your plan's cap (4 hours at most).

Failures distinguish regional or licensing restrictions, temporary download failures, and unsupported content. Retry a temporary download failure later. If the source restricts access, try another source or upload a file you can access.