Text this: Content-Adaptive Inference for State-of-the-Art Learned Video Compression