Lectures become sources.
Scribe v2 transcribes recordings with word timings and speakers. Passages break at a new speaker, a pause, or a sentence end past 45 seconds, so a citation plays the recording from its first word.
listening and speaking
Your sources,
out loud.
Konkyo turns a space into an audio brief: two ElevenLabs voices talk through its sources, and every fact they say is cited. Scribe transcribes your lectures and hears you when you talk to it.
step into the library Make a brief from any space: + add, then audio brief.interactive exampleelevenlabs × konkyo
caffeine, sleep, and memory
0:00 / 1:35
Bella · line 1 of 11
Today we’re looking at how caffeine acts in the brain, and what it may mean for sleep and memory.
an opening line: no fact, no citation
A real brief Konkyo made from four papers on caffeine: two voices from Text to Dialogue, each line timed by Scribe, each fact cited.
Follow one brief.
From one small action to what you see,
each step opened in the code.
- 1
The Writer scripts two hosts.
A short conversation about the space's sources, each fact with its quote, checked in Zetaris like any answer.
- 2
Text to Dialogue speaks it.
eleven_v3 turns the whole script into one recording, in two voices.
- 3
Scribe times every line.
The dialogue's own timings drift up to 1.8 s after 30 s, so Scribe listens to the finished audio and each line is timed from its words.
- 4
The player follows along.
The line being said lights up with its source; tap its citation and the passage opens.
ElevenLabs in KonkyoWhere it
Where it
shows up.
Specific jobs in the app.
Each one has a place you can point to.
Talk to it.
The orb listens with Scribe v2 Realtime in your browser, primed with words from your library so titles and names come out right.
Answers, read aloud.
Flash reads an answer in the voice you picked, with word timings for the highlight. A clip is made once and kept, so playing it again costs nothing.
Say it, in study.
Practice you can answer out loud: Scribe hears you, and your answer is checked like a typed one.
A closer look.
The choices behind the behaviour.
Open whichever part you're curious about.
Your key stays on the server
The browser listens with a single-use Scribe token made for that session, never the API key.
Words to passages
Timed words are grouped into passages at a speaker change, a 2.5 s pause, or once a passage passes its length and reaches a sentence end.
Two hosts, two voices
Briefs use eleven_v3 at 128 kbps with two fixed voices; guests make one a day, accounts five.
Daily caps
Speaking uses Flash v2.5. Each person gets 80 voice sessions and 8,000 spoken characters a day (guests 12 and 1,500); a cached clip doesn't count.