Ejemplo n.º 1
0
    def compose(self, score, **optional):
        params= bind_params(self.params, optional)
        
        melody_instrument= None
        rythm_instrument= None
        for instrument in score.instruments:
            if self.matches_description(instrument, params['melody_patch'], params['melody_channel']):
                melody_instrument= instrument
            if self.matches_description(instrument, params['rythm_patch'], params['rythm_channel']):
                rythm_instrument= instrument
        if (melody_instrument is None or rythm_instrument is None) and len(score.instruments) > 1: raise Exception("Que instrument?")
        else: 
            rythm_instrument= melody_instrument= score.instruments[0]
        
        interval_size= measure_interval_size(score, params['n_measures']) 
        
        #harmonic_context_alg= YamlHarmonicContext('/home/prakuso/tesis/src/electrozart/electrozart/composers/base2.yaml', score.divisions)
        harmonic_context_alg= ChordHarmonicContext(score)
        harmonic_context_alg= PhraseRepetitions(harmonic_context_alg, alpha=7)

        rythm_alg= RythmHMM(interval_size, instrument=rythm_instrument.patch, channel=rythm_instrument.channel)
        phrase_rythm_alg= rythm_alg
        phrase_rythm_alg= ListRythm(rythm_alg)
        phrase_rythm_alg= RythmCacheAlgorithm(ListRythm(rythm_alg), 'phrase_id')

        melody_alg= NarmourHMM(instrument=melody_instrument.patch, channel=melody_instrument.channel)
        phrase_melody_alg= melody_alg
        phrase_melody_alg= ListMelody(melody_alg)
        phrase_melody_alg= CacheAlgorithm(ListMelody(melody_alg), 'phrase_id')

        #rythm_score= score.copy()
        #rythm_score.notes_per_instrument.pop(piano)
        print "todos los intrsumentos"
        for instrument in score.instruments:
            if not instrument.is_drums:
                melody_alg.obsSeqBuilder.builder.patch= instrument.patch
                melody_alg.train(score)
            rythm_alg.obsSeqBuilder.builder.patch= instrument.patch
            rythm_alg.train(score)

        harmonic_context_alg.train(score)

        applier= AlgorithmsApplier(harmonic_context_alg, phrase_rythm_alg, phrase_melody_alg)
        self.applier= applier
        applier.start_creation()
        rythm_alg.draw_model('rythm.png', score.divisions)
        melody_alg.model.draw('melody.png', str)

        duration= score.duration
        #duration= harmonic_context_alg.harmonic_context_alg.chordlist[-1].end
        #duration= harmonic_context_alg.chordlist[-1].end
        
        # octave range
        score_notes= score.get_notes(skip_silences=True)
        mean_pitch= sum(float(n.pitch) for n in score_notes)/len(score_notes)
        std_dev= sqrt(sum((n.pitch-mean_pitch)**2 for n in score_notes)/len(score_notes))
        #import ipdb;ipdb.set_trace()
        octave= int(mean_pitch/12) #+ 1
        min_pitch= octave*12
        max_pitch= (octave+2)*12 + 6
        offset= 0
        min_pitch= int(mean_pitch - std_dev+offset)
        max_pitch= int(mean_pitch + std_dev+offset)

        print "MIN PITCH", min_pitch
        print "MAX PITCH", max_pitch
        general_input= AcumulatedInput()
        general_input.min_pitch= min_pitch
        general_input.max_pitch= max_pitch


        notes= applier.create_melody(duration, params['print_info'], general_input=general_input)
        #seed(time())

        #chord_notes= []
        #for c in harmonic_context_alg.chordlist:
        #    for n in c.notes:
        #        chord_notes.append(PlayedNote(n.pitch+3*12, c.start, c.duration, 80))
        #duration= chord_notes[-1].end

        res= score.copy()
        rythm_alg.model.calculate_metrical_accents()
        #rythm_alg.model.draw_accents('accents.png', score.divisions)
        import random
        cache= {}
        def random_accent(note):
            moment= (note.start%rythm_alg.model.interval_size)/rythm_alg.model.global_gcd
            res= cache.get(moment)
            if res is None:
                res= randint(1, 6)
                cache[moment]= res
            return res
        def dec_accent(note):            
            moment= (note.start%rythm_alg.model.interval_size)/rythm_alg.model.global_gcd
            res= cache.get(moment)
            if res is None:
                res= 7-moment
                cache[moment]= res
            return res
        def inc_accent(note):            
            moment= (note.start%rythm_alg.model.interval_size)/rythm_alg.model.global_gcd
            res= cache.get(moment)
            if res is None:
                res= moment + 1 
                cache[moment]= res
            return res

        accent_func= rythm_alg.model.get_metrical_accent 
        accent_func= inc_accent
        accent_func= dec_accent
        accent_func= random_accent

        #seed(time())
        #max_accent= max(rythm_alg.model.metrical_accents.itervalues())
        #min_accent= max(rythm_alg.model.metrical_accents.itervalues())
        max_accent= 6
        for i, n in enumerate(notes):
            accent= accent_func(n)
            n.volume=  min(100, max(60, (100 * accent)/max_accent))
            if not n.is_silence: n.pitch+=12
            #if rythm_alg.model.get_metrical_accent(n)== min_accent  and random.random() > 0.5:
            #    pass
            #    #notes[i]= Silence(n.start, n.duration)

        print "min volume:", min(n.volume for n in notes if not n.is_silence) 
        print "max volume:", max(n.volume for n in notes if not n.is_silence)

        print "accent mapping"
        for moment, accent in sorted(cache.items(), key=lambda x:x[0]):
            print moment, accent
        for n in res.get_notes(skip_silences=True):
            n.volume= 50

        #piano= res.notes_per_instrument.keys()[0]
        #piano= Instrument()
        instrument= Instrument()
        instrument.patch= params['melody_patch_to_dump']
        instrument.patch= 26
        instrument.patch= 73
        instrument.patch= 21
        instrument.patch= 74
        res.notes_per_instrument[instrument]= notes
        #res.notes_per_instrument= {instrument: notes, melody_instrument:res.notes_per_instrument[melody_instrument]}
        #res.notes_per_instrument= {instrument: notes}
        #res.notes_per_instrument[piano]= chord_notes

        #rythm_alg.draw_model('rythm.png')

        import electrozart
        from electrozart import base
        from electrozart.algorithms import ExecutionContext
        #from pycana import CodeAnalyzer

        #analyzer= CodeAnalyzer(electrozart)
        #relations= analyzer.analyze(exceptions= [ExecutionContext])
        #analyzer.draw_relations(relations, 'relations.png')


        return res
Ejemplo n.º 2
0
    def compose(self, score, **optional):
        params= bind_params(self.params, optional)
        
        piano= score.instruments[0]
        interval_size= measure_interval_size(score, params['n_measures']) 
        
        harmonic_context_alg= YamlHarmonicContext('/home/prakuso/tesis/src/electrozart/electrozart/composers/base34.yaml', score.divisions)
        harmonic_context_alg= PhraseRepetitions(harmonic_context_alg)

        rythm_alg= RythmHMM(interval_size, multipart=False, instrument=piano.patch, channel=piano.channel)
        phrase_rythm_alg= RythmCacheAlgorithm(ListRythm(rythm_alg), 'part_id')
        #phrase_rythm_alg= rythm_alg

        melody_alg= NarmourHMM(instrument=piano.patch, channel=piano.channel)
        #phrase_melody_alg= ListMelody(melody_alg)
        phrase_melody_alg= CacheAlgorithm(ListMelody(melody_alg), 'part_id')
        #phrase_melody_alg= melody_alg

        rythm_alg.train(score)
        harmonic_context_alg.train(score)
        melody_alg.train(score)

        applier= AlgorithmsApplier(harmonic_context_alg, phrase_rythm_alg, phrase_melody_alg)

        duration= harmonic_context_alg.harmonic_context_alg.chordlist[-1].end

        notes= applier.create_melody(duration, params['print_info'])

        chord_notes= []
        for c in harmonic_context_alg.harmonic_context_alg.chordlist:
            for n in c.notes:
                chord_notes.append(PlayedNote(n.pitch+3*12, c.start, c.duration, 80))
        duration= chord_notes[-1].end

        instrument= Instrument()
        instrument.patch= params['melody_instrument']

        res= score.copy()
        rythm_alg.model.calculate_metrical_accents()
        rythm_alg.model.draw_accents('accents.png', score.divisions)
        max_accent= max(rythm_alg.model.metrical_accents.itervalues())
        import random
        for i, n in enumerate(notes):
            accent= rythm_alg.model.get_metrical_accent(n)
            n.volume=  max(62, (100 * accent)/max_accent)
            if not n.is_silence: n.pitch+=24
            if accent == 1: #  and random.random() > 0.5:
                pass
                #notes[i]= Silence(n.start, n.duration)

        for n in res.get_notes(skip_silences=True):
            n.volume= 70

        piano= res.notes_per_instrument.keys()[0]
        piano= Instrument()
        instrument.patch= 74
        instrument.patch= 73
        res.notes_per_instrument= {}
        res.notes_per_instrument[instrument]= notes
        res.notes_per_instrument[piano]= chord_notes

        #rythm_alg.draw_model('rythm.png')

        return res