File size: 790 Bytes
9ed1a09
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
db8eebf
 
9ed1a09
 
 
dda0117
 
9ed1a09
 
 
 
 
 
 
 
 
 
 
 
db8eebf
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
---
tags:
- text-to-image
- lora
- diffusers
- template:diffusion-lora
widget:
- output:
    url: images/pixel_composite.png
  text: Make the sea shell red
- output:
    url: images/pixel_composite.png
  text: '-'
base_model: FreedomIntelligence/Janus-4o-7B
instance_prompt: null
license: mit
datasets:
- osunlp/MagicBrush
---
# ThinkClearly

https://github.com/peterN908/ThinkClearly/tree/main


## Model description 

A conceptual exploration of tool use in AR image models 

We fine-tune Janus 4o models to produce `<CLEAR>` instruction tokens when making text+image to image edits. This indicate that this token patch should be an exact pixel match to the input image



## Download model


[Download](/peterface/ThinkClearly/tree/main) them in the Files & versions tab.