Posts

Showing posts from May, 2024

TikTok Marketing Report using ai!

 I was curious to try this out. I wanted to plug my social media data into chatgpt and see if it could provide suggestions for me in order to boost engagement. Pretty neat! I would like to give this another try with more data because I used only a few days worth of data where I only posted once or twice a day. I only used a few days worth of data because I had to turn this feature on in TikTok and was impatient. Check out the brief report below. TikTok Marketing Report Period: April 14 - April 20, 2024 Overview of Account Performance: Total Video Views: 1,094 Total Reached Audience: 648 Total Profile Views: 12 Total Likes: 87 Total Comments: 7 Engagement Rate: 8.59% Detailed Video Performance Analysis: Videos posted from April 17 to April 20, including times, engagement metrics, and content types, indicating an active posting schedule towards the end of the week. Highest engagement seen in videos tagged with #bluegrass and #flatpicking, showing a specific interest in these genres from

Automatic 1111 Stable Diffusion (September 29th, 2023)

Image
Automatic 1111 Stable Diffusion September 29th, 2023 I installed automatic 1111 stable diffusion. I never experimented with the sampling methods until now. I didn’t expect the image to change completely when changing the sampling method, even though I kept the same seed. Some sampling methods produce similar or identical results such as PLMS, DPM++ 2M, LMS Karras, and DPM2 Karras. I experimented with different CFG scales too and actually referred to the model I used called absolute reality . The details for that model said to use cfg 4.5-10 but I actually liked the result I got from a cfg scale of 3. I put the cfg scale at the highest (30) and the image came out bad, the shapes weren’t well defined and it didn’t have much detail.  I started with the default sampling steps at 20 and then tried the maximum. The maximum sampling steps came out nice so I kept my sampling step at 150.  Something that I learned from your last lecture is to always check the notes from any model I use. I didn’

Trying out AnimeGAN (september 23rd, 2023)

Image
  AnimeGAN September 23rd animeGAN_inference I spent a lot of time trying to locally install animeGAN on my computer. What kept me from using animeGAN on my computer, is that I got an error related to CUDA not being compatible with the version of python I had. I’ve run into CUDA issues before and I still don’t quite understand the point of CUDA and why different programs require different CUDA versions and also why GPU’s require additional CUDA installations. The error message was suggesting that I install an older python version. I decided not to proceed because I was afraid it would mess up my local installation of a different AI tool that I use on a weekly basis. I tried to figure out how to make a virtual environment for animeGAN but I still got the same error. I looked into setting up a virtual environment because the other AI image based tool I use, uses a “venv”. Since I got the same CUDA error when trying to set up a virtual environment, I decided it’s probably best I hold off

Exploring AI Art Generation with Midjourney (October 8th, 2023)

Image
  October 8th, 2023 Exploring AI Art Generation through Image Variations and Stylistic Experiments I focused on exploring the creative possibilities of Midjourney, an AI image generation tool, by experimenting with different prompts and settings to produce variations of an image depicting "Ronald McDonald." My objective was to create a series of images portraying Ronald McDonald's family, including his wife and daughter. Through trial and error, I adjusted the parameters, particularly the prompt strength, to achieve a realistic photograph-like appearance. Despite requiring multiple attempts, this experimentation resulted in high-quality, realistic images that aligned well with my vision. Experiment 1: Ronald McDonald’s Family The initial focus was to develop realistic photographic depictions of Ronald McDonald’s wife and daughter. The challenge was balancing prompt strength, number of inference steps, and guidance scale to yield a photograph-like aesthetic. The following