<h3 id="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.-exnb7ptoz">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.</h3><p>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.</p><p>Training data used for this particular LoRA were dated <strong>1994 through 1999</strong>.</p><p>Dataset preparation using random segments of video were prepared using scripts from my github - <a target="_blank" rel="ugc" href="https://github.com/akspa0/dataset-tools/blob/main/video/randomSampler/videosampler2x6.py">https://github.com/akspa0/dataset-tools/blob/main/video/randomSampler/videosampler2x6.py</a></p>

Files with SHA256: a490d9a98f5adb44e9ac6c1fe675e85e75f7ec36fae13c2474234a9b87eb1598