SerSleepy on Nostr: so turns out higgsfield was too expensive for this, i usually use it for 5-9 sec ...

so turns out higgsfield was too expensive for this, i usually use it for 5-9 sec videos
——
train LoRa for character consistency & specialized shots

& use stable video diffusion models or Wan 2.2 with vid2vid

this would be experimental, getting that desired output of resembling a waking life is harder than imagined