Ideas Scratchpad
Last backup: 04-Oct 2018
1.Crowd sourced sounds. Synthesize music with these sounds matching some target. All sounds collected at the same time across the world
2. Make music where progressively the entire tuning rises until it’s almost a semitone higher till the end. And then it stops down to the original tuning. Will this make it sound weird or resolving?
3. Make a multi-dimensional visual for seeing quality of singing. Similar to TELMI project work on violin sounds quality.
4. Make a physical pouch of fluid (air?) That can be touched for tactile music. (Similar to jacket in concert effect)
5. Take a picture, select objects that are then identified and tracked. Then each unique object can be a sample trigger that is set when tapped on the screen. Animating regular objects as instruments (subsidiary idea : make music with balls in front of camera where ball corresponds to a control as another instrument)
6. Jokes idea: a web based space where you can listen to a chosen song; but in different environments (like in any concert from the back, through the telephone, radio, etc. To study environmental effects on sound moulding (modified hrtf of an entire crowd, for the audience at the back)
7. Deep learning to model from waveform to spectrogram. To understand what are the learning steps upto spectrogram, instead of looking at an end to end approach
8. A piano app (or other instrument) that opens up on top of background music playing in the phone (like on Spotify or SoundCloud) – Found app!
9. Audio in painting. Remove a part and try to generate it back and measure it’s Turing test
10. Phonation mode: gives quality of singing voice (app style assisting for singers)
11. Jacobify a song. Add crazy changes to an existing song (maybe in midi)? (Todo: think some more what this might mean, study Jacob Collier changes)
12. Harmonizer instrument for guitar fingers. Jacob type instrument for the guitar (really want to do this pet project. Initial research shows, requirement of MIDI guitar or making a hex pickup. Able to generate vocoder style voicings)
13. Fun project: jacobify – create voicing parameter and choose some anchor chords and then randomly use plagal plagal cadence (with 4ths and 5ths and make bass part chromatic) to generate an arrangement
14. Music idea: (Forest orchard Forest) two grooves with one of them slowly getting out of synch with the other groove. And the quantized parts in every division (8 7 6 5 4 3 2 ) or (8 7 4 3 6 5 2) is emphasized with chords/voicings on different instrument. (Check shpongle dmt )
15. Use last fm dataset to identify how many people care about lyrics of a song (check if it has timing based like/favorite) could definitely be done by Spotify dataset
16. Classifying albums by album art
17. Musical idea: a groove idea overlaps with a little phase difference. (Like a Canon – row row)
18. Musical idea:1) Mandelbrod set mapping to sound
2) auralizing other variants of the mandelbrod set
19. Sensor idea: measure the direction and speed the phone is moving in and use it to visualize/auralize something. (Maybe while traveling in a car or metro one could feel the motion of the metro)
20. Metronome that says syllables/ Indian rhythmic solefege instead of clicks (ta ki ta) etc. Also puzzle like where you can segment pleases together to create groove – konnakol/bol
21. Jazzify a viral video clip, like mononeon. Capture pitch arcs, in human speech. (Notes – Bass follows pitch arc, adding synth to make interesting chords, regular repetition of a speech sequence makes musical illusions)
22. Piano transcription for vocal recording (almost black midi) (like the audio illusion). A program that takes vocal part and backing track and converts vocals to crazy piano part creating the vocal illusion)
23. Jokes idea – Emojify a chord progression. – change the cadence into a string of emojis. Or a video of my face getting happy, sad, confused etc.
24. NLP sentence description for a given sound. (From freesound – nlp for audio scene analysis). (Sub problem -Automatic tag creation from sound.)
25. (Not related to my work, check for toilets with urine and poop testing for health in normal households, quick way of testing common conditions. Important in poorer localities without access to regular tests.)
26. Visual Concept idea for art: Man with square shaped head trying to fit in a circle and at the same time man with circle shape trying to get in square. Concept: everyone is wearing each other’s face
27. Genetic algo idea: a) convert one fixed melody into another fixed melody using genetic algorithms.
b) natural selection : website has various algorithm generated music and people can vote it as interesting or not (scoring), those that survive mix to create more favorable melodies. (Note: this might be easier to score as a comparison of two generated melodies to choose a preference)
28. Human voice instrument: a synth where you can control the vocal tract and breathiness along with playing a melody.
29. A plug-in to make a sound richer, by resynthesizing several versions of the same sound (changing timbre in every version slightly) to create a group texture. Ex- record a clap/vocal “oo” sound and make it variably richer. (Probably works best with voice, because it’s a known instrument of many variations but worth trying on other sounds)
30. An emotional chord maker – choose only notes and then over a 2d affective emotion grid choose the emotion of the chord voicing. (App perhaps? Or plugin)
31. Jacob type: Quantifying the voice leading in a chord progression. (Giving preference to chromatic movement of parts and unchanged parts to suggest alternate chord voicing for a progression)
32. Composing tool, where you sing or play melody and choose when a chord should appear only by choosing the color of brightness you want it to be. (On a color wheel perhaps)
33. Dancing animoji – a visual avatar that dances to the music and shows emotions (emoji) along with some music. (Possible browser version to watch with music)
34. Braille music! Research the standard and try to make Braille notation for more musical expression and playback
35. Generating doodles (from quick draw dataset), given a sentence (nlp, to break down key words). So converting a sentence from a poem into meaningful string of visual doodles.
36. Visualization idea : A website (AR ?) procedural visual that follows the music (visual idea : paper planes flying in circles to rhythm getting added with colors based on chords, perspective change with sectio change, chorus identification)
37. Generate music and visuals for spoken word/text/poem. Example – adding music to a speech recording making it musical.
Example – text gets meter based on syllables. Visual elements related to the subject/noun etc. (Word net nlptk)
38. Visual idea (art): hand connected with strings to another hand. Each controls the other.
39. Show your hand on camera and move your fingers, each finger controls a limb of a puppet on screen. (Visual project/ AR project)
40. Visualization project : circle of fifths visual for any given music. (Mir to obtain key and chords and beats) ex – John Coltrane (Untitled song) video on my playlist
41. Given a Melody (Twinkie twinkle) automatic reharmonize generator (with parameter for crazy timing/swing/ density, oddity, simplicity)
42. Make a word cloud for every decade of hip hop/pop lyrics (top albums from billboard) to determine changing trends of music lyrics writing
43. Physical sounds waves in the space around us (AR) moving your phone through the space to hear it. (Permanence as against ephemeral sound)
44. Jokes idea : given any freesound clip, generate the lick out of it.
45. Analysis of all songs, take lakh midi dataset and analyze for histograms of popular chord structures, repetitions, length of Melody, hook (take ideas of musical structure from Andrew huang’s crowd funded Melody project)
46. transcribing audiation (Melody in my head) through EEG signals. (Check literature to see if possible: most literature was about eeg signals when analysing heard pitch and not while audiating a pitch)
47. Detection (and classification) of visual onsets. Cues that could be onsets (in video) to be corresponding onsets or transients in the accompanying audio. Scoring automation for a video. Maybe as an added device in video editing tools where it quickly shows all the repetitive and onset regions in the video to automatically add/suggest scoring options.
48. Geo tagged visualization on a map of various 1) music tech labs 2) field wise papers in ismir and other conferences. (P5.js)
49. Singing voice synthesis, given a segment of speech and a melody line, synthesize singing voice. (Yamaha’s vocaloid. Also check voctrolabs startup with similar product. demo works for a fixed melody)
50. Tala/drum cycles generated from famous number series. (Check exhaustive number series online). Perhaps a measure of the musicality/musical beauty of each series.
51. Musical idea: chromatic connected lines moving in space (like a spatial motion, like left to right in stereo) experiment with multiple lines in spatial journeys. (Example the different lines in a Bach 4 voice chorale)
52. Harmonic/melodic emotional palette : a browser based survey kind of thing where you can choose from some emotions (and add textual notes to exemplify) and then it arranges the harmonic and melodic options in a heirarchy. (Note: I can use this to collect my emotional thoughts about each choice)
53. Browser app: choose a song from lakh midi dataset, choose a sample from freesound. Playback entire song with the chosen sample. This can be done from sample picked from mic input as well.
54. Point phone camera at symbolic score and it plays the corresponding midi notes at a fixed tempo
55. Music idea: swinging of percussive section controlled by heart rate onsets. (Optical music recognition)
56. Vocalizer, synthesized singing a melody of Indian classical symbolic sequence (in a particular raag) singing a vowel sound. (Using raag rules to make chalans, pakkads). There are small datasets with notations for common indian raag phrases. Vocalizing them with a vowel sound instead of a sinusoid sample.
57. Speech command to a mobile device that gives result in an ear piece. (SciFi type project example can do math calculations in ear discreetly)
58. Aligning word2vec embeddings for song/album/genre reviews to the LSTM embeddings of the song midis. To learn pairwise relations. This could generate midi from description or generate description from midi. (Might need lot of data to learn abstract relationships, too many variants. Think about a controlled dataset for testing this idea, single reviewer, or single band)
59. Deep learning for recreating simple math equations like addition, multiplication (does bodmas heirarchy get captured in various layers/nodes) and more complicated mappings of series expansion coefficients etc.
60. Song summarizing by a dynamic/structure arc (visual) and a 2 second texture of the song.
61. Paper classification by tags. (Try on ismir abstracts)
62. fun project: face solo: mapping a face eyebrows, scowl, head movement to guitar solo notes and phrases or just to music
63. 2dimensional db spl vs hz range area plots for music by genre, decade, artist. For insight.
64. Crazy idea: record and playback sound corresponding to video but by reversing the stereo channel. (Equivalent to functioning with prism glasses) does the mind learn to invert for localization of sound like it does in visual cortex.
65. Composing phrase lines with emoji sequences.
66. Music idea: compose a song where the tension and release arcs follow the stress and easy years of my life. So, a song whose arcs are representations of my journey (tension is great, but only when your release it, remember to do that in life)
Manaswi Mishra 