VOICES IN MY HEAD · ANTON MATZKAIM

wav length · LOOP makes a file that repeats with no seam

live · who is talking

press TALK: complete phrases and replies appear here

1who speaks

NATURAL: real voices recorded ahead of time. MY VOICE: phrases you record in step 5. ROBOT: the old speech chip. pick one or mix them.

each voice separately · new people · seed

seed

2what they say

edit the lines · write your own

one line per row. a line no set holds is spoken by the ROBOT voice, the natural voices only know the ready sets. for a dialogue: one turn per row, two speakers taking turns, an empty row starts a new scene.

3how they talk

CROWD: many people at once, now and then one clear voice. RELAY: they speak in turn around your head. SOLO: one voice that jumps around you.

presets · a ready sound in one press

4main knobs

5my voice

say what you will record, press REC, speak, press again. every phrase is kept in this browser and the crowd can say it.

mic off

microphone · import files

IMPORT AUDIO FILES

voice memos from your phone work too: each file becomes one phrase named after the file.

record the browser's own voices

the voices of your system cannot be routed into the page, so they are caught from the speakers and kept as your phrases. three steps, each one says what is wrong.

1 · a voice that speaks

looking for voices

2 · an input that hears the speakers

3 · speak every line of the text box and keep it

no voices, or only robots, on Linux: install RHVoice, four free Russian voices for every browser, then restart the browser: sudo apt install rhvoice rhvoice-russian speech-dispatcher-rhvoice. Windows has Microsoft Svetlana and Dmitry, a Mac has Milena and Yuri.

all knobs · crowd, words, space, throat fx, face, room

how the voices are made

LINES → PEOPLE → CROWD / RELAY / SOLO → WHOLE / CUT-UP / BABBLE → STUTTER · BACKWARDS · WARP → THROAT FX · more the further → FLIGHT AROUND YOUR HEAD → ROOM → DUCK UNDER THE CLEAR VOICE → SQUEEZE → speakers
VOICES TALKING → INSANITY → YOUR FACE + SIGIL

the default throats are natural voices baked ahead of time: every line and word of the ready sets was spoken once by the free Silero and Piper engines and kept as a small file, so they speak the moment the set downloads. a line you type yourself, which no set holds, is spoken by the SAM chip instead. MY VOICE plays phrases you recorded. five instant SAM voices stay for the robotic sound; neural voices are the slow backup and load only when picked. the system voices of your browser cannot be used: the browser plays them past the page, no effect can reach them. whole phrases keep their complete audio; word slicing is reserved for deliberate CUT-UP and BABBLE effects. Pairs take turns in a scripted conversation while the background crowd keeps talking. a scheduler gives each speaker a place around you, a path to fly while it talks, a shade that darkens it behind your head and far away, and a send into the room that grows with distance. on the radar: you in the middle facing up, squares are where people stand, dots are voices talking, red is closer than a metre. the face in the middle is the head the voices are in: it tears and melts under the voices talking right now, doubles in red towards the ones closest to it, and grows a sigil once the room gets loud