I just wanted to comment that the technology and process behind this video is truly amazing. As usual, I generated about 100 images of Mia to start with, then I picked these 7 images to create a video. The incredible part is that these images are the "keyframes" - the start and beginning of each segment of video.
With the current video Ai models I have, I usually generate less than 5-8 second videos from an image but the quality or actions start to degrade after the 8 second mark. This keyframe method I'm using now works a lot better so hopefully I'll be able to do better videos in the future with higher quality.
I also discovered that using a video transition called "Cross Dissolve" of 4 frames will smooth out the transitions between each generated video. It's subtle but very effective.
Everything (including Mia's voice) was Ai generated. I put about 4 hours of pretty hard work that mostly was editing and finetuning. The Ai models and my computers did the rest.
I might be selling myself a little short. It does take a LOT of creativity, choices and input to create these videos, it's hard to rely on the Ai model to just randomly generate an action that is good. I do spend time choosing the music and "directing" the video through the images I generate and choose. I also finetune the "voice" of the character but it's a hit and miss process.
I guess it does help that I have a bit of real experience with video editing and film production, but I definitely don't have "professional" experience. If you've been on this Patreon for a while, you'll definitely see how the videos have evolved over time.
I'm curious to know what everyone thinks. Do you feel this video is "different" from the past ones. Is it better/worse than what you've seen here so far? I'm biased because I'm so immersed in the effort and process of creating so it's really difficult for me to judge my own work objectively. I'm excited because I know I hit a technical milestone with this video and not necessarily an artistic one. I'm trying to keep my videos under 2 min now due to time and effort.
Liana Heward
2025-05-28 19:10:05 +0000 UTCLeigh jones
2025-05-28 18:59:13 +0000 UTC