Skip to content

commands.md

Latest commit

 

History

History
118 lines (99 loc) · 4.22 KB

commands.md

File metadata and controls

118 lines (99 loc) · 4.22 KB
Raw

oram — command grammar

all commands, whether keyboard, parsed text, or LLM-assisted, become a structured action before touching the engine.

parser priority

  1. normalize text: lowercase, strip punctuation, collapse whitespace.
  2. extract layer references: one→1, two→2, three→3, four→4.
  3. extract durations: seconds first, then bars if BPM is available.
  4. match deterministic regex/rules.
  5. if no match and LLM is enabled, ask the LLM for one allowed action schema.
  6. if still uncertain, do not execute. show "unrecognized command" in TUI.

poetic input is allowed, but output must be structured and bounded.

transport commands

record eight seconds       -> record(duration=8.0)
record 8 seconds           -> record(duration=8.0)
record four bars           -> record(bars=4) (requires BPM)
stop recording             -> stop_recording()
kill audio                 -> kill_audio()
loop this                  -> (set mode to loop)
overdub                    -> overdub(target=selected)
mute layer two             -> mute_layer(target=2)
solo layer one             -> solo_layer(target=1)
clear layer one            -> request clear confirmation for layer 1
clear layer one (repeat)   -> clear_layer(target=1)
select layer three         -> select_layer(target=3)
save scene                 -> save_session()
export mix                 -> export_mix()

mix commands

set volume layer 1 0.50    -> set_volume(target=1, volume=0.5)
volume layer two 50        -> set_volume(target=2, volume=0.5)
set pan layer 3 -0.75      -> set_pan(target=3, pan=-0.75)
pan layer one left         -> set_pan(target=1, pan=-0.75)
pan layer two right        -> set_pan(target=2, pan=0.75)
pan layer two center       -> set_pan(target=2, pan=0.0)
make everything softer     -> reduce active layer volumes by 20%

transform commands

reverse layer one          -> apply_effect(target=1, effect=reverse)
make layer one slower      -> apply_effect(target=1, effect=speed, speed=0.5)
make it slower             -> apply_effect(target=selected, effect=speed, speed=0.5)
pitch it down              -> apply_effect(target=selected, effect=pitch, semitones=-2)
pitch layer two up five    -> apply_effect(target=2, effect=pitch, semitones=5)
fade the end               -> apply_effect(target=selected, effect=fade_out)
trim the beginning         -> apply_effect(target=selected, effect=trim_start)
filter the voice           -> apply_effect(target=selected, effect=lowpass)
make it darker             -> apply_effect(target=selected, effect=lowpass)
make it thinner            -> apply_effect(target=selected, effect=highpass)
granulate softly           -> apply_effect(target=selected, effect=granular, density=0.3)
stretch it until it breathes -> apply_effect(target=selected, effect=stretch_breathe)

spatial commands

move it far away           -> apply_effect(effect=spatial_far)
turn this into a small room -> apply_effect(effect=reverb, decay=short)
add more distance          -> apply_effect(effect=spatial_far)
make the room narrower     -> apply_effect(effect=reverb, narrow=true)
wash it in reverb          -> apply_effect(effect=reverb, wet=0.8)

generative commands

add distant metallic rain  -> generate_layer(prompt="distant metallic rain")
create a room tone         -> generate_layer(prompt="room tone")
summon a synthetic forest  -> generate_layer(prompt="synthetic forest")
add a low unstable drone   -> generate_layer(prompt="low unstable drone")
make a fake field recording -> generate_layer(prompt="fake field recording")
generate a quiet machine   -> generate_layer(prompt="quiet machine ambience")

listening commands

listen to the texture      -> analyze_mix()
describe the current mix   -> analyze_mix()
what is too dense          -> analyze_mix(focus="density")
what changed in layer two  -> analyze_mix(target=2)
find speech residue        -> analyze_mix(focus="speech")
analyze the loop           -> analyze_mix()

action types

core action types:

  • record
  • stop_recording
  • kill_audio
  • overdub
  • select_layer
  • mute_layer
  • solo_layer
  • clear_layer
  • set_volume
  • set_pan
  • apply_effect
  • remove_effect
  • generate_layer
  • analyze_mix
  • save_session
  • export_mix
  • set_mode
  • quit