
Include the year you want to target in your prompt, like 1994, 1998, or 1999, and either 'tv' or 'commercial', to ensure that the model gets included in the output.
Trained against frames from tv commercials recorded onto VHS from my youtube channel, @Dadsweathertapes. Frames were upscaled from SD footage and retained haloing, chroma issues, and other VHS artifacts. The VHS tapes were digitized in the last 2 years for the above youtube channel, using a standard 4-head consumer VCR and a Collossus II input card.
Training data used for this particular LoRA were dated 1994 through 1999.
Dataset preparation using random segments of video were prepared using scripts from my github - https://github.com/akspa0/dataset-tools/blob/main/video/randomSampler/videosampler2x6.py
描述:
训练词语:
名称: 837665_training_data.zip
大小 (KB): 199275
类型: Training Data
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
名称: retroVHS1994-1999.safetensors
大小 (KB): 18908
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success