Commit ·
92d685b
1
Parent(s): 1209876
Update README.md
Browse files
README.md
CHANGED
|
@@ -4,11 +4,15 @@ base_model:
|
|
| 4 |
---
|
| 5 |
|
| 6 |
This is FireRed Image Edit, which combines Transformer Diffusion, Text Encoder, and VAE into a single file.
|
| 7 |
-
|
| 8 |
-
This model is still unstable for *Text to Images* (T2I) at resolutions of 1280x720 and above.
|
| 9 |
-
However for image editing, the results are more stable and satisfactory.
|
| 10 |
|
| 11 |
*Version 1*
|
| 12 |
- Combined With LoRa 8 Steps From FireRed Repo
|
| 13 |
- MMProj stays in FP16
|
| 14 |
-
- CFG 1.0 | Euler-A / Beta, Normal, Simple | 8 Steps
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
---
|
| 5 |
|
| 6 |
This is FireRed Image Edit, which combines Transformer Diffusion, Text Encoder, and VAE into a single file.
|
| 7 |
+
For some reason FireRed 1.1 models (F16 or GGUF) have difficulty editing high resolution PNG images, JPG format is stable.
|
|
|
|
|
|
|
| 8 |
|
| 9 |
*Version 1*
|
| 10 |
- Combined With LoRa 8 Steps From FireRed Repo
|
| 11 |
- MMProj stays in FP16
|
| 12 |
+
- CFG 1.0 | Euler-A / Beta, Normal, Simple | 8 Steps
|
| 13 |
+
|
| 14 |
+
*Version 2*
|
| 15 |
+
- Combined With LoRa 4/8 Steps Qwen-Image-2512
|
| 16 |
+
- CFG 1.0 |
|
| 17 |
+
- Recommeded Sampler : Euler_a,
|
| 18 |
+
- 4 Steps for Fast, 8 Steps for better Quality
|