Instructions to use Lightricks/LTX-2-Marking with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- AudioSeal
How to use Lightricks/LTX-2-Marking with AudioSeal:
# Watermark Generator from audioseal import AudioSeal model = AudioSeal.load_generator("Lightricks/LTX-2-Marking") # pass a tensor (tensor_wav) of shape (batch, channels, samples) and a sample rate wav, sr = tensor_wav, 16000 watermark = model.get_watermark(wav, sr) watermarked_audio = wav + watermark# Watermark Detector from audioseal import AudioSeal detector = AudioSeal.load_detector("Lightricks/LTX-2-Marking") result, message = detector.detect_watermark(watermarked_audio, sr) - Notebooks
- Google Colab
- Kaggle
File size: 940 Bytes
718f58d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 | checkpoint_path: https://dl.fbaipublicfiles.com/videoseal/pixelseal/checkpoint.pth
args:
attenuation_config: configs/attenuation.yaml
attenuation: jnd_1_1
nbits: 256
hidden_size_multiplier: 1.0
img_size_proc: 256
blending_method: additive
scaling_w: 0.2
scaling_i: 1.0
videowam_chunk_size: 32
videowam_step_size: 8
embedder:
model: unet_base_yuv_quant
params:
msg_processor:
nbits: 16
hidden_size: 32
msg_processor_type: 'binary+concat'
unet:
in_channels: 1
out_channels: 1
z_channels: 16
num_blocks: 8
activation: 'relu'
normalization: 'batch'
z_channels_mults: [2, 4, 8, 16]
last_tanh: True
extractor:
model: convnext_tiny
params:
encoder:
depths: [3, 3, 9, 3]
dims: [96, 192, 384, 768]
pixel_decoder:
pixelwise: False
upscale_stages: [1]
embed_dim: 768
nbits: 16
sigmoid_output: False
|