A project built out of interest in a show that gives you something different every time you watch it — in different years, at different points in your life.
It started as an archive. The journal, the music, and the characters took it somewhere else.
It opens with Back in the 90’s, a cue that sets the narrative context for everything that follows.
It has an archive of episodes, characters, and quotes. It has AI-powered conversations with five main characters, each with their own ambient sound. It has AI-generated voices — hear what the characters might say next, in their tone, on your topic. It has a journal with music tracks, each paired with its own set of writing prompts — questions drawn from the emotional register of the music and the questions the show keeps asking.
what's inside
- Characters — Main cast and recurring figures, with descriptions and context.
- Episodes — All six seasons, with summaries and season filtering.
- Quotes — Organized by season, fully searchable.
Voices and Talk exist because the show ended and the characters didn't stop feeling real.
Talk is not a chatbot. It's a room. You walk in, the ambient sound starts, and someone who feels like them is there. You can ask BoJack what he thinks about something that happened to you today. You can hear what Diane would say. The character prompts and the ambient sound exist because a room needs to feel like a room, not a feature.
Voices is for the episodes that don't exist yet. The show ended beautifully and completely — it shouldn't continue, and it won't. But you can still hear what they might say next. About your topic, in their voice, in whatever tone fits the day.
BoJack will make you laugh before he says something that lands harder than you expected. Diane will take your question seriously, maybe too seriously. Princess Carolyn will answer three things at once and somehow address all of them. Mr. Peanutbutter will be so genuinely warm you forget he's not listening. Todd will say something that doesn't quite make sense and then make complete sense.
the rooms
When you first open Talk, before you choose a character, there is a room. A quiet, generic hum — the sound of being somewhere without being anywhere in particular. It fades as soon as you step in.
Each character has their own sound, built entirely with the Web Audio API. No recordings. Every sound is synthesized:
- BoJack Horseman — a big empty house at 2am. A TV on somewhere in another room. A low melancholic drone in E, with a slow vibrato, and a distant upper harmonic that gives it that hollow, unbothered quality. The sound of someone who has been alone long enough that it stopped feeling like anything.
- Diane Nguyen — a Chicago apartment. Rain against the window, filtered into that specific soft hiss of water on glass. Street noise below, low and muffled. A radiator hum underneath it all. Warmth that costs something to maintain.
- Princess Carolyn — a Hollywood office after hours. HVAC running. Fluorescent lights at high frequency. Perpetual motion with no one left to witness it.
- Mr. Peanutbutter — sunny Bel Air, open and warm. A G major chord sustained in the air — four harmonics with slow vibrato — like someone who decided a long time ago that the world was good and has never stopped believing it.
- Todd Chavez — wherever Todd is. Mid-range room noise, soft and neutral. Irregular piano pings in C major, arriving at random intervals, never announcing themselves. The sound of someone who is always slightly somewhere else.
the journal
A private writing space. Music tracks, each tuned to a different emotional register. Prompts shift based on what you're listening to. If the music doesn't match how you hear it, free shuffle pulls from all moods with no interpretation.
The journal has two ways to listen to each track — a synthesized version and a Spotify stream. The synth is built entirely with the Web Audio API. No audio files, no samples. Each melody is transcribed note by note from the original recording and rendered using sine wave oscillators with a handcrafted piano envelope. It runs entirely in your browser. Spotify streams the real thing.
Music and mood mean different things to different people. These pairings are one interpretation — yours may differ.
how the sound is built
The ambient sound doesn't try to be realistic. It tries to feel like distance crossed.
Time march — you are moving through time to reach them. Not a live conversation. Something that requires crossing a distance that isn't just space.
Unrealistic — you know it isn't real. That's part of it. The knowing makes it more tender, not less. You are choosing to believe for a moment. Like holding a transistor radio to your ear and listening for a signal from somewhere that doesn't broadcast anymore.
Old radio — the static, the imperfection, the analog warmth. Not a clean digital connection. Something that carries the feeling of distance and time in the sound itself. The noise is the honesty of it. It says — this is far away. This took effort to reach.
Each character's room is synthesized from scratch — oscillators and filtered noise, built the way analog radio builds sound from electromagnetic waves. Imperfect by nature. Warm because of the imperfection. You are not opening a chat window. You are traveling somewhere.
The foundation of every room is white noise — a buffer of random values covering all frequencies equally. By itself it's harsh and bright. The character of each room comes from what gets removed. A lowpass filter cuts everything above a certain frequency. What remains is the low, soft rumble — the bottom of the sound. From there each character gets their own layers. Oscillators for specific tones — BoJack's TV hum at 180hz, Diane's radiator at 110hz, Princess Carolyn's fluorescent at 3960hz.
Every sound fades in slowly rather than snapping on. You don't notice it arriving — it's just suddenly there.
the journal synth
The journal's synthesized tracks are built the same way — entirely in-browser, no audio files. Each melody is transcribed note by note from the original recording and stored as an array of frequency and duration pairs. The synth plays those exact pitches in the exact rhythm of the original.
Every note is shaped with a handcrafted piano envelope — attack rises from silence in 30ms, sustain holds through the middle of the note, release fades to near-zero before the next note begins. The oscillator type is sine, the purest wave, close to the tone of a piano's upper register without being harsh.
What it can't do: no overtones, no pedal resonance, no dynamics. Every note hits at the same velocity. It doesn't try to sound like a real piano. It sounds like someone humming the melody to themselves — the shape of the music, quietly present while you write. That's why it works.
They are, honestly, a little cute. A simple ascending melody in pure sine tone, playing softly while you write. That combination feels like something.
For the real thing, Spotify is one button away.