While many of us have gradually reduced our lockdown activities, online innovator Teigan Reamsbottom has taken a unique approach by waking up at 4 a.m. to curate classic pop culture movie clips. Utilizing a specialized processing unit filled with RAM and software employing machine-learning techniques, he endeavors to enhance the original characteristics of the videos to the fullest extent possible.
The discovery of pop-culture relics is indeed heartening, even amidst a landscape where films from the 1980s and 1990s may risk being lost due to tax considerations and the limitations of traditional media.
One notable example is a scene from the iconic film “Sex and the City,” featuring Kim Cattrall singing while her then-husband Mark plays an upright bass. This video has long been a staple on the internet and was even showcased in a THNK1994 exhibition at a gallery in the Lower East Side. It epitomizes Reamsbottom’s vision for the delightfully eccentric and sometimes cringe-worthy camp aesthetic of a bygone era. Its revival in a specific corner of the internet sparked widespread celebration. Contrasting the restored version with the vague clip we have been accustomed to watching for years:
However, the process of expanding videos to enhance their quality for modern high-resolution displays is fraught with challenges. Each technique employed leaves behind artifacts that impact the visual appeal of the new footage, making the data capture and conversion a labor-intensive task. As reliance on machine-learning models grows, so does the risk of unintended consequences.
Professional 4K transfers may sometimes appear overly polished or artificial, detracting from the original charm of the content. AI-powered video software attempts to extrapolate missing details based on patterns, often resulting in inaccuracies. Despite its shortcomings, this approach is deemed essential by Reamsbottom for preserving moments that might otherwise fade into obscurity.
Experimenting with various AI models and adjusting parameters like sharpness and shadows is a continuous process for Reamsbottom to mitigate the common pitfalls of upscaling programs. The AI algorithms may inadvertently exaggerate details like teeth, creating unnatural effects that require careful fine-tuning to rectify.
The meticulous nature of this work cannot be overstated. Even with dedicated hardware, processing a 30-minute video can consume more than a day. Achieving optimal results from pixelated or low-quality source material poses additional challenges, demanding extensive expertise and patience.
While the realm of high-quality video restoration remains primarily within the purview of industry professionals, advancements in generative AI technology are making tools accessible to enthusiasts willing to invest time and resources. Reamsbottom relies on Topaz Labs’ software suite, a $300 investment, to achieve satisfying outcomes from his collaborations, such as the restoration of Connie Francis tapes provided by a fan’s family.
The transformation of these archival materials through deep learning models evokes a sense of rediscovery and emotional connection, underscoring the profound impact of visual enhancement efforts. The painstaking process yields results that breathe new life into cherished memories, resonating with viewers on a deeply personal level.
