| |
| |
| |
|
|
| import os |
| import time as reqtime |
| import datetime |
| from pytz import timezone |
|
|
| import gradio as gr |
|
|
| import numpy as np |
|
|
| import os |
| import random |
| from collections import Counter |
| import math |
| from tqdm import tqdm |
|
|
| import TMIDIX |
|
|
| from midi_to_colab_audio import midi_to_colab_audio |
|
|
| |
|
|
| def Generate_Chords_Progression(minimum_song_length_in_chords_chunks, |
| chords_chunks_memory_ratio, |
| chord_time_step, |
| merge_chords_notes, |
| melody_MIDI_patch_number, |
| chords_progression_MIDI_patch_number, |
| base_MIDI_patch_number |
| ): |
|
|
| print('=' * 70) |
| print('Req start time: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now(PDT))) |
| start_time = reqtime.time() |
|
|
| print('=' * 70) |
| print('Requested settings:') |
| print('Minimum song length in chords chunks:', minimum_song_length_in_chords_chunks) |
| print('Chords chunks memory ratio:', chords_chunks_memory_ratio) |
| print('Chord time step:', chord_time_step) |
| print('Merge chords notes max time:', merge_chords_notes) |
| print('Melody MIDI patch number:', melody_MIDI_patch_number) |
| print('Chords progression MIDI patch number:', chords_progression_MIDI_patch_number) |
| print('Base MIDI patch number:', base_MIDI_patch_number) |
| print('=' * 70) |
| |
| |
|
|
| print('=' * 70) |
| print('Pitches Chords Progressions Generator') |
| print('=' * 70) |
|
|
| print('=' * 70) |
| print('Chunk-by-chunk generation') |
| print('=' * 70) |
| print('Generating...') |
| print('=' * 70) |
| |
| matching_long_chords_chunks = [] |
| |
| ridx = random.randint(0, len(all_long_chords_tokens_chunks)-1) |
| |
| matching_long_chords_chunks.append(ridx) |
| |
| max_song_len = 0 |
| |
| tries = 0 |
| |
| while len(matching_long_chords_chunks) < minimum_song_length_in_chords_chunks: |
|
|
| matching_long_chords_chunks = [] |
| |
| ridx = random.randint(0, len(all_long_chords_tokens_chunks)-1) |
| |
| matching_long_chords_chunks.append(ridx) |
| seen = [ridx] |
| gseen = [ridx] |
| |
| for a in range(minimum_song_length_in_chords_chunks * 10): |
| |
| if not matching_long_chords_chunks: |
| break |
| |
| if len(matching_long_chords_chunks) > minimum_song_length_in_chords_chunks: |
| break |
| |
| schunk = all_long_chords_tokens_chunks[matching_long_chords_chunks[-1]] |
| trg_long_chunk = np.array(schunk[-chunk_size:]) |
| idxs = np.where((src_long_chunks == trg_long_chunk).all(axis=1))[0].tolist() |
| |
| if len(idxs) > 1: |
| |
| random.shuffle(idxs) |
| |
| eidxs = [i for i in idxs if i not in seen] |
| |
| if eidxs: |
| eidx = eidxs[0] |
| matching_long_chords_chunks.append(eidx) |
| seen.append(eidx) |
| gseen.append(eidx) |
| |
| if 0 < chords_chunks_memory_ratio < 1: |
| seen = random.choices(gseen, k=math.ceil(len(gseen) * chords_chunks_memory_ratio)) |
| elif chords_chunks_memory_ratio == 0: |
| seen = [] |
| |
| else: |
| gseen.pop() |
| matching_long_chords_chunks.pop() |
| |
| else: |
| gseen.pop() |
| matching_long_chords_chunks.pop() |
| |
| |
| if len(matching_long_chords_chunks) > max_song_len: |
| print('Current song length:', len(matching_long_chords_chunks), 'chords chunks') |
| print('=' * 70) |
| final_song = matching_long_chords_chunks |
| |
| max_song_len = max(max_song_len, len(matching_long_chords_chunks)) |
| |
| tries += 1 |
| |
| if tries % 500 == 0: |
| print('Number of passed tries:', tries) |
| print('=' * 70) |
|
|
| if len(matching_long_chords_chunks) > max_song_len: |
| print('Current song length:', len(matching_long_chords_chunks), 'chords chunks') |
| print('=' * 70) |
| final_song = matching_long_chords_chunks |
| |
| f_song = [] |
| |
| for mat in final_song: |
| f_song.extend(all_long_good_chords_chunks[mat][:-chunk_size]) |
| f_song.extend(all_long_good_chords_chunks[mat][-chunk_size:]) |
| |
| print('Generated final song after', tries, 'tries with', len(final_song), 'chords chunks and', len(f_song), 'chords') |
| print('=' * 70) |
|
|
| print('Done!') |
| print('=' * 70) |
| |
| |
| |
| print('Rendering results...') |
| print('=' * 70) |
|
|
| output_score = [] |
| |
| time = 0 |
| |
| patches = [0] * 16 |
| patches[0] = chords_progression_MIDI_patch_number |
| |
| if base_MIDI_patch_number > -1: |
| patches[2] = base_MIDI_patch_number |
| |
| if melody_MIDI_patch_number > -1: |
| patches[3] = melody_MIDI_patch_number |
| |
| chords_labels = [] |
| |
| for i, s in enumerate(f_song): |
| |
| time += chord_time_step |
| |
| dur = chord_time_step |
| |
| chord_str = str(i+1) |
| |
| for t in sorted(set([t % 12 for t in s])): |
| chord_str += '-' + str(t) |
| |
| chords_labels.append(['text_event', time, chord_str]) |
| |
| for p in s: |
| output_score.append(['note', time, dur, 0, p, max(40, p), chords_progression_MIDI_patch_number]) |
| |
| if base_MIDI_patch_number > -1: |
| output_score.append(['note', time, dur, 2, (s[-1] % 12)+24, 120-(s[-1] % 12), base_MIDI_patch_number]) |
| |
| if melody_MIDI_patch_number > -1: |
| output_score = TMIDIX.add_melody_to_enhanced_score_notes(output_score, |
| melody_patch=melody_MIDI_patch_number, |
| melody_notes_max_duration=max(merge_chords_notes, chord_time_step) |
| ) |
|
|
| if merge_chords_notes > 0: |
| escore_matrix = TMIDIX.escore_notes_to_escore_matrix(output_score) |
| output_score = TMIDIX.escore_matrix_to_merged_escore_notes(escore_matrix, max_note_duration=merge_chords_notes) |
| |
| midi_score = sorted(chords_labels + output_score, key=lambda x: x[1]) |
|
|
| fn1 = "Pitches-Chords-Progression-Composition" |
| |
| detailed_stats = TMIDIX.Tegridy_ms_SONG_to_MIDI_Converter(midi_score, |
| output_signature = 'Pitches Chords Progression', |
| output_file_name = fn1, |
| track_name='Project Los Angeles', |
| list_of_MIDI_patches=patches |
| ) |
| |
| new_fn = fn1+'.mid' |
| |
| |
| audio = midi_to_colab_audio(new_fn, |
| soundfont_path=soundfont, |
| sample_rate=16000, |
| volume_scale=10, |
| output_for_gradio=True |
| ) |
| |
| |
|
|
| output_midi_title = str(fn1) |
| output_midi = str(new_fn) |
| output_audio = (16000, audio) |
| |
| output_plot = TMIDIX.plot_ms_SONG(output_score, plot_title=output_midi, return_plt=True) |
| |
| print('Done!') |
| print('=' * 70) |
| |
| |
|
|
| print('Generated chords progression info and stats:') |
| print('=' * 70) |
|
|
| chords_progression_summary_string = '=' * 70 |
| chords_progression_summary_string += '\n' |
|
|
| all_song_chords = [] |
| |
| for pc in f_song: |
| tones_chord = tuple(sorted(set([p % 12 for p in pc]))) |
| all_song_chords.append([pc, tones_chord]) |
| |
| print('=' * 70) |
| print('Total number of chords:', len(all_song_chords)) |
| chords_progression_summary_string += 'Total number of chords: ' + str(len(all_song_chords)) + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| print('=' * 70) |
| print('Most common pitches chord:', list(Counter(tuple([a[0] for a in all_song_chords])).most_common(1)[0][0]), '===', Counter(tuple([a[0] for a in all_song_chords])).most_common(1)[0][1], 'count') |
| chords_progression_summary_string += 'Most common pitches chord: ' + str(list(Counter(tuple([a[0] for a in all_song_chords])).most_common(1)[0][0])) + ' === ' + str(Counter(tuple([a[0] for a in all_song_chords])).most_common(1)[0][1]) + ' count' + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| print('=' * 70) |
| print('Most common tones chord:', list(Counter(tuple([a[1] for a in all_song_chords])).most_common(1)[0][0]), '===', Counter(tuple([a[1] for a in all_song_chords])).most_common(1)[0][1], 'count') |
| chords_progression_summary_string += 'Most common tones chord: ' + str(list(Counter(tuple([a[1] for a in all_song_chords])).most_common(1)[0][0])) + ' === ' + str(Counter(tuple([a[1] for a in all_song_chords])).most_common(1)[0][1]) + ' count' + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| print('=' * 70) |
| print('Sorted unique songs chords set:', len(sorted(set(tuple([a[1] for a in all_song_chords])))), 'count') |
| chords_progression_summary_string += 'Sorted unique songs chords set: ' + str(len(sorted(set(tuple([a[1] for a in all_song_chords]))))) + ' count' + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| |
| for c in sorted(set(tuple([a[1] for a in all_song_chords]))): |
| |
| chords_progression_summary_string += str(list(c)) + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| print('=' * 70) |
| print('Grouped songs chords set:', len(TMIDIX.grouped_set(tuple([a[1] for a in all_song_chords]))), 'count') |
| chords_progression_summary_string += 'Grouped songs chords set: ' + str(len(TMIDIX.grouped_set(tuple([a[1] for a in all_song_chords])))) + ' count' + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| print('=' * 70) |
| for c in TMIDIX.grouped_set(tuple([a[1] for a in all_song_chords])): |
| |
| chords_progression_summary_string += str(list(c)) + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| |
| |
| chords_progression_summary_string += 'All songs chords' + '\n' |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| |
| for i, pc_tc in enumerate(all_song_chords): |
| |
| chords_progression_summary_string += 'Song chord # ' + str(i) + '\n' |
| |
| chords_progression_summary_string += str(list(pc_tc[0])) + ' === ' + str(list(pc_tc[1])) + '\n' |
| |
| chords_progression_summary_string += '=' * 70 |
| chords_progression_summary_string += '\n' |
| |
| |
| print('-' * 70) |
| print('Req end time: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now(PDT))) |
| print('-' * 70) |
| print('Req execution time:', (reqtime.time() - start_time), 'sec') |
|
|
| return output_audio, output_plot, output_midi, chords_progression_summary_string |
|
|
| |
|
|
| if __name__ == "__main__": |
| |
| PDT = timezone('US/Pacific') |
| |
| print('=' * 70) |
| print('App start time: {:%Y-%m-%d %H:%M:%S}'.format(datetime.datetime.now(PDT))) |
| print('=' * 70) |
|
|
| |
|
|
| soundfont = "SGM-v2.01-YamahaGrand-Guit-Bass-v2.7.sf2" |
|
|
| print('Loading processed Pitches Chords Progressions dataset data...') |
| print('=' * 70) |
| long_tones_chords_dict, all_long_chords_tokens_chunks, all_long_good_chords_chunks = TMIDIX.Tegridy_Any_Pickle_File_Reader('processed_chords_progressions_chunks_data') |
| |
| print('=' * 70) |
| print('Resulting chords dictionary size:', len(long_tones_chords_dict)) |
| print('=' * 70) |
| print('Loading chords chunks...') |
|
|
| chunk_size = 4 |
| |
| src_long_chunks = np.array([a[:chunk_size] for a in all_long_chords_tokens_chunks]) |
| |
| print('Done!') |
| print('=' * 70) |
| print('Total chords chunks count:', len(all_long_good_chords_chunks)) |
| print('=' * 70) |
| |
| |
| |
| app = gr.Blocks() |
| with app: |
| gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>Chords Progressions Generator</h1>") |
| gr.Markdown("<h1 style='text-align: center; margin-bottom: 1rem'>Generate unique chords progressions</h1>") |
| gr.Markdown( |
| "\n\n" |
| "This is a demo for Tegridy MIDI Dataset\n\n" |
| "Check out [Tegridy MIDI Dataset](https://github.com/asigalov61/Tegridy-MIDI-Dataset) on GitHub!\n\n" |
| "[Open In Colab]" |
| "(https://colab.research.google.com/github/asigalov61/Tegridy-MIDI-Dataset/blob/master/Chords-Progressions/Pitches_Chords_Progressions_Generator.ipynb)" |
| " for all options, faster execution and endless generation" |
| ) |
|
|
| gr.Markdown("## Select generation options") |
|
|
| minimum_song_length_in_chords_chunks = gr.Slider(4, 60, value=30, step=1, label="Minimum song length in chords chunks") |
| chords_chunks_memory_ratio = gr.Slider(0, 1, value=1, step=0.1, label="Chords chunks memory ratio") |
| chord_time_step = gr.Slider(100, 1000, value=500, step=50, label="Chord time step") |
| merge_chords_notes = gr.Slider(0, 4000, value=1000, step=100, label="Merged chords notes max time") |
| melody_MIDI_patch_number = gr.Slider(0, 127, value=40, step=1, label="Melody MIDI patch number") |
| chords_progression_MIDI_patch_number = gr.Slider(0, 127, value=0, step=1, label="Chords progression MIDI patch number") |
| base_MIDI_patch_number = gr.Slider(0, 127, value=35, step=1, label="Base MIDI patch number") |
|
|
| run_btn = gr.Button("generate", variant="primary") |
|
|
| gr.Markdown("## Generation results") |
| |
| output_audio = gr.Audio(label="Output MIDI audio", format="mp3", elem_id="midi_audio") |
| output_plot = gr.Plot(label="Output MIDI score plot") |
| output_midi = gr.File(label="Output MIDI file", file_types=[".mid"]) |
|
|
| output_cp_summary = gr.Textbox(label="Generated chords progression info and stats") |
|
|
| run_event = run_btn.click(Generate_Chords_Progression, |
| [minimum_song_length_in_chords_chunks, |
| chords_chunks_memory_ratio, |
| chord_time_step, |
| merge_chords_notes, |
| melody_MIDI_patch_number, |
| chords_progression_MIDI_patch_number, |
| base_MIDI_patch_number], |
| [output_audio, output_plot, output_midi, output_cp_summary] |
| ) |
| |
| app.queue().launch() |
