<h3 id="include-the-year-you-want-to-target-in-your-prompt-like-1986-1987-1992-or-1993-and-either-'tv'-or-'commercial'-to-ensure-that-the-model-gets-included-in-the-output.-9su5dvk0p">Include the year you want to target in your prompt, like 1986, 1987, 1992 or 1993, and either 'tv' or 'commercial', to ensure that the model gets included in the output.</h3>Use 0.5 to 1.0 strengthTrained 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 1986 through 1993.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>

Files with SHA256: 6db073e8cab40aad7d06da2b5f6ae40f40cc6199fea6f12b885cffd3d836eec9