Our great sponsors
-
InfluxDB
Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.
One way to avoid the inpainting issue is to use Ddetailer or Adetailer to auto-inpaint detected faces. I use the latter as I couldn't get ddetailer to work, but the fundamental feature is the same. It does a face-detection and inpainting pass after the initial generation (and hires fix), which means your first output already has the face fixed somewhat. And then you can go in to correct smaller details.
One way to avoid the inpainting issue is to use Ddetailer or Adetailer to auto-inpaint detected faces. I use the latter as I couldn't get ddetailer to work, but the fundamental feature is the same. It does a face-detection and inpainting pass after the initial generation (and hires fix), which means your first output already has the face fixed somewhat. And then you can go in to correct smaller details.
After months of wrangling with Dreambooth, I finally mastered how to use it. The training rate is the key. I found a spreadsheet on the Dreambooth webui extension github discussion forum. I don't know if most people are aware of it. For me, it has been extremely reliable. In fact, I think the formulas in it should be built into Dreambooth trainers.
Related posts
- WHY IS THIS HAPPENING MAKE IT STOP I HATE IT
- Launch HN: Rubbrband (YC W23) – Deformity detection for AI-generated images
- Automatically skip GFPGAN on faces that are "too low quality"
- How to diffuse better faces?
- Is there anything which automatically recognizes when something was ill-generated (such as a face, or fingers etc.) and automatically applies the inpainting mask to that area?