i)ddlmZ ddlmZddlmZdZddl Z ddl Z ddl m Z ddl mZddlZ ddlmZmZdd lmZdZdd lmZmZddlZdd lmZmZmZmZm Z m!Z!dd l"mZm#Z#dd l$m%Z%ddl&m'Z'dde#deeddfdZ( dde#deede)de*de+de,f dZ- dde%de)de*de+de,deee+ff dZ.y#e$rdZYwxYw#e$rdZYwxYw))Union)Audio)VideoTFN)Path) AudioClip VideoClip)mplfig_to_npimage)MappingOptional) AnnotationSegment SlidingWindowSlidingWindowFeatureTimelinenotebook)r AudioFile)Model)Binarize audio_filesegmentreturnctstjdy|td|\}}ntdj ||\}}t |j |S)adlisten to audio Allows playing of audio files. It will play the whole thing unless given a `Segment` to crop to. Parameters ---------- audio_file : AudioFile A str, Path or ProtocolFile to be loaded. segment : Segment, optional The segment to crop the playback to. Defaults to playback the whole file. z-You need IPython installed to use this methodNdownmix)mono)rate)IPYTHON_INSTALLEDwarningswarnrcrop IPythonAudioflatten)rrwaveformsrs q/Volumes/fast/ai/experiments/voice-extract-mac/.venv/lib/python3.12/site-packages/pyannote/audio/utils/preview.pylistenr%Csb  EF,u),Z8 "),11*gF " ((* 44zoom video_fps video_extdisplayc tstjdy|rtstjdt |t r d|vr|d}n7t |t tfrt|j}n tdtjd}|d|d |} td d } | j|} td | | j|\j!j"t%j&dz dz } t)| t+j,t+j.fdt0ffd } t3| j4}t7j8t;dz}t=j>|ddd|zfd\}|dddf^ddt0ff d }tA|j4}|jC|}|jE| |dddt=jF|s| StI| d S)!aPreview Parameters ---------- audio_file : AudioFile A str, Path or ProtocolFile to be previewed segment : Segment, optional The segment to crop the preview to. Defaults to preview the whole file. video_fps : int, optional Video frame rate. Defaults to 5. Higher frame rate leads to a smoother video but longer processing time. video_ext : str, optional One of {"webm", "mp4", "ogv"} according to what your browser supports. Defaults to "webm" as it seems to be supported by most browsers (see caniuse.com/webm)/ display : bool, optional Wrap the video in a IPython.display.Video instance for visualization in notebooks (default). Set to False if you are only interested in saving the video preview to disk. **views : dict Additional views. See Usage section below. Returns ------- * IPython.display.Video instance if `display` is True (default) * path to video preview file if `display` is False Usage ----- >>> assert isinstance(annotation, pyannote.core.Annotation) >>> assert isinstance(scores, pyannote.core.SlidingWindowFeature) >>> assert isinstance(timeline, pyannote.core.Timeline) >>> preview("audio.wav", reference=annotation, speech_probability=scores, speech_regions=timeline) # will create a video with 4 views. from to bottom: # "reference", "speech probability", "speech regions", and "waveform") z-You need MoviePy installed to use this methodNzSince IPython is not installed, this method cannot be used with default display=True option. Either run this method in a notebook or use display=False to save video preview to disk.urizUnsupported 'audio_file' type.zpyannote-audio-preview)prefix/.i>r) sample_raterg)startend)r1durationstepTct|tjr6tj|zj tj St |zSN) isinstancenpndarraytakeastypeint64round)r6datar0s r$make_audio_framez!preview..make_audio_framesI a $774!k/!9!9"((!CD DE!k/*++r&)r4fps F)nrowsncolsfigsizesqueezer)gg?cN |jz}t|d zz |d zzt_j tj d j ||g djjjdjjjgjdt jD]+\\}}}|j t|trtj ||dnSt|t"rtj$||ddn)t|t&rtj ||d |j ||g d|jjjd|jjjg|jd j)|j+d .t-S) Ng?T)axtimeylimzk--r"F)rJrK)rJrKlegend _)r1r rrclear plot_featureplotaxes get_yaxis set_visible set_ticks set_ylabelzipitemsr9r plot_timeliner plot_annotationrjoinsplitr )r6tnameviewax_viewax_viewsax_wavfigrviewsr"rL ylim_waveformr's r$ make_framezpreview..make_frames  C$JC$J?  h6=Q  QFM51  ++D1 ))"-*%%(%A !LT4' MMO$)&&teDD*-(('dSD"67%%dwUN LL!Qu - LL " " $ 0 0 6 LL " " $ . .r 2   sxx 38 9'&B*!%%r&)r4T ultrafastbar)rBaudio audio_fpspresetlogger)embed)%MOVIEPY_INSTALLEDrrrr9r strrr_ ValueErrortempfilemkdtempr get_durationr rnumpyr6rr1rr:minmaxfloatrr4rresetlenpltsubplotsr set_audiowrite_videofileclose IPythonVideo)rrr'r(r)r*rer,temp_dir video_pathrjr4samplesrA audio_cliprErSrg video_cliprbrcr@rdr0r"rLrfs `` ` @@@@@@@@r$previewr\s2b  EF(  M *g&5J+> Jd ,:##9::'?@H:Qse1YK0J e) 4E%%j12 :w7D+ ::<>>Dmma+oA OG$D'2HFF4L"&&,.M,E, +g6F6FKXJ NN JNE 1r1u9ouICQT Xv D'&e'&'&R:0@0@AJ%%j1J  IIcN   $ //r&modelblankctt|j}|dj\}}}|dj\}} } |jj } t |||| zzz} t |j| d} tjd| f}tj| | f}tt|d|dD]L\}\}}t |||| zzz}||dd|||zf<|j|\}}||||| zddf<Ntt||jj}|| dd}t|||||S) aPreview training samples of a given model Parameters ---------- Model : Model Model, already setup for training (i.e. call model.setup(stage="fit") first). blank : float, optional Add blank of that many seconds between each sample. Defaults to 1.0 video_fps : int, optional Video frame rate. Defaults to 5. Higher frame rate leads to a smoother video but longer processing time. video_ext : str, optional One of {"webm", "mp4", "ogv"} according to what your browser supports. Defaults to "webm" as it seems to be supported by most browsers (see caniuse.com/webm)/ display : bool, optional Wrap the video in a IPython.display.Video instance for visualization in notebooks (default). Set to False if you are only interested in saving the video preview to disk. Returns ------- * IPython.display.Video instance if `display` is True (default) * path to video preview file if `display` is False Xyrr3NTrainingSamples)r"r0r,)r(r)r* reference)nextitertrain_dataloadershaperjr0int introspectiontorchzeros enumeraterXrrframesr)rrr(r)r*batch batch_size num_channels num_samples num_frames num_speakersr0batch_num_samplesbatch_num_framesr"rbrrX_idxy_idxrOrs r$BROKEN_preview_training_samplesrsD e,,./ 0E,1#J,<,<)J k+0:+;+;(J L++))KJ+ 8K*KLM5../@A!DE{{A012H -|<=Is5:uSz:; 6AqAu{'::;<34EEK///0&&u-q34 %%*,,a/0 <  / 5;N;N;U;UVWI" J   r&r8)Ng$@webmT)g?rrT)/typingrIPython.displayrr rrr ImportErrorrrrpathlibrmatplotlib.pyplotpyplotr{rur:moviepy.editorrrmoviepy.video.io.bindingsr ror r r pyannote.corer r rrrrpyannote.audio.core.iorpyannote.audio.core.modelrpyannote.audio.utils.signalrr%rxrrpboolrrr&r$rs^0553; % 4+05y58G+<556"& d0d0 g d0 d0 d0  d0  d0R @ @ @@ @  @  <  @Is"B1B>1B;:B;>CC