lllyasviel / lllyasviel/ControlNet

Generated Images are not like the input one

Open
#287 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
34.1k
Forks
3k
PR merge metrics
No merged PRs in 30d

Description

Dear great authors,

Currently I've tried to use canny info as the input of ControlNet.
My used case is a child portrait. As the code described, the input will generate edge info and combined with prompts to de-noise sequentially.
Finally, it can generate a child based on the scribble and edge info. However, the generated child does not look like the input child
Is it possible the results can be more similar to the input image? That is, preserve the identity characters? Thanks

PS: I've tried to alter guidance-scale/strength already

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

The issue names no files, tests, or entry points; start by reproducing the child-portrait example with Canny input and the reported guidance-scale and strength settings. Done would require an agreed way to measure identity preservation and results that are more similar to the input, but the issue does not define either.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.