o ۷i@s4ddlmZddlmZddlmZddlZddlZ ddl Z ddl m Z ddl m Z mZe eZeZer:ddlZnedd ejjd ed ejjfd d Zd ejjdejjdejd dfddZd ejjdejjde jd ed df ddZ ddee jjej Be jBee jBdede jd ede!ded dfddZ"dS))Iterator)Fraction)chainN)tqdm) get_loggeris_av_availablez`PyAV is required to use LTX 2.0 video export utilities. You can install it with `pip install av` containeraudio_sample_ratereturncCs0|jd|d}||j_d|j_td||j_|S)z/ Prepare the audio stream for writing. aacratestereo) add_stream codec_context sample_ratelayoutr time_base)r r audio_streamr[/home/ubuntu/vllm_env/lib/python3.10/site-packages/diffusers/pipelines/ltx2/export_utils.py_prepare_audio_stream(s rrframe_inc Cs|j}|jpd}|jp d}|jp|j}tjjj|||d}d}||D]} | j dur.|| _ || j 7}|j| _| | | q$| D]} | | qDdS)Nfltpr)formatrrr) rrrravaudio resamplerAudioResamplerresampleptssamplesmuxencode) r rrcc target_format target_layout target_rateaudio_resampleraudio_next_ptsrframepacketrrr_resample_audio3s&       r.r#cCs|jdkr |dddf}|jddkr|jddkr|j}|jddkr.td|jd|jtjkrCt|dd}|dtj}t j j | dd d d d }||_t|||dS) Nrrz,Expected samples with 2 channels; got shape .gg?g@s16r)rr)ndimshapeT ValueErrordtypetorchint16cliptor AudioFrame from_ndarray contiguousreshapecpunumpyrr.)r rr#r rrrr _write_audioPs  rBrvideofpsr output_pathvideo_chunks_numbercCst|tr"t|dtjjr"dd|D}tj|dd}t|}n/t|tjrQt t ||k|t |k}t |rG|d d}ntdt|}t|tjrctj||dd}t|}t|}|j\} } } } tj|d d } | jd t|d } | | _| | _d | _|dur|durtdt| |}tt |g||ddD]#}|!d"}|D]}tj#j$|dd}| %|D]}| &|qqq| %D]}| &|q|durt'| |||| (dS)a$ Encodes a video with audio using the PyAV library. Based on code from the original LTX-2 repo: https://github.com/Lightricks/LTX-2/blob/4f410820b198e05074a1e92de793e3b59e9ab5a0/packages/ltx-pipelines/src/ltx_pipelines/utils/media_io.py#L182 Args: video (`List[PIL.Image.Image]` or `np.ndarray` or `torch.Tensor`): A video tensor of shape [frames, height, width, channels] with integer pixel values in [0, 255]. If the input is a `np.ndarray`, it is expected to be a float array with values in [0, 1] (which is what pipelines usually return with `output_type="np"`). fps (`int`) The frames per second (FPS) of the encoded video. audio (`torch.Tensor`, *optional*): An audio waveform of shape [audio_channels, samples]. audio_sample_rate: (`int`, *optional*): The sampling rate of the audio waveform. For LTX 2, this is typically 24000 (24 kHz). output_path (`str`): The path to save the encoded video to. video_chunks_number (`int`, *optional*, defaults to `1`): The number of chunks to split the video into for encoding. Each chunk will be encoded separately. The number of chunks to use often depends on the tiling config for the video VAE. rcSsg|]}t|qSr)nparray).0framerrr sz encode_video..)axisuint8zSupplied `numpy.ndarray` does not have values in [0, 1]. The values will be assumed to be pixel values in [0, ..., 255] and will be used as is.)dimw)modelibx264r yuv420pNz4audio_sample_rate is required when audio is providedzEncoding video chunks)totaldescr@rgb24)r)) isinstancelistPILImagerGstackr8 from_numpyndarray logical_and zeros_like ones_likeallroundastypeloggerwarningTensor tensor_splititernextr4ropenrintwidthheightpix_fmtr6rrrr;rA VideoFramer=r%r$rBclose)rCrDrr rErF video_framesis_denormalized first_chunk_rmrlr streamr video_chunkvideo_chunk_cpu frame_arrayrJr-rrr encode_videonsN           ry)r)#collections.abcr fractionsr itertoolsrrArG PIL.ImagerYr8rutilsrr__name__rd _CAN_USE_AVr ImportErrorr Containerrkr AudioStreamrr<r.rfrBrXrZr]strryrrrrsh        $