Card: samples section, ComfyUI node link
Browse files
README.md
CHANGED
|
@@ -79,6 +79,28 @@ Anima Pixel Head Tap before the KSampler and uses Anima Pixel Decode in place of
|
|
| 79 |
|
| 80 |
ComfyUI keeps the VAE on the card after its first decode, so the VAE column has no move cost.
|
| 81 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 82 |
Held-out reads during training (dB, colour / mono):
|
| 83 |
|
| 84 |
| decode point | colour | mono |
|
|
@@ -99,7 +121,8 @@ Held-out reads during training (dB, colour / mono):
|
|
| 99 |
|
| 100 |
## Use
|
| 101 |
|
| 102 |
-
- ComfyUI: the ComfyUI-Anima-PixelDecoder
|
|
|
|
| 103 |
add **Anima Pixel Head Tap** on the MODEL before the KSampler (stock or Spectrum) and decode
|
| 104 |
the sampler's LATENT with **Anima Pixel Decode** instead of VAE Decode. Run the DiT on the
|
| 105 |
fp32 stream (Anima Adapter Loader, `stream = fp32`).
|
|
|
|
| 79 |
|
| 80 |
ComfyUI keeps the VAE on the card after its first decode, so the VAE column has no move cost.
|
| 81 |
|
| 82 |
+
## Samples
|
| 83 |
+
|
| 84 |
+
Turbo 4 steps, CFG 1, seed 42, 1024×1024, in ComfyUI (the workflow above); left VAE Decode,
|
| 85 |
+
right Anima Pixel Decode, decoded from the same latent. Prompt: `masterpiece, best quality,
|
| 86 |
+
score_7, safe. An anime girl wearing a black tank-top and denim shorts is standing outdoors.
|
| 87 |
+
She's holding a rectangular sign out in front of her that reads "ANIMA". She's looking at the
|
| 88 |
+
viewer with a smile. The background features some trees and blue sky with clouds.`; the
|
| 89 |
+
monochrome pair adds `monochrome, comic` after `safe`. Crops: 384×384 at native resolution,
|
| 90 |
+
x 192–576, y 512–896.
|
| 91 |
+
|
| 92 |
+
| | VAE Decode | pixel head |
|
| 93 |
+
|---|---|---|
|
| 94 |
+
| colour, 37.3 dB |  |  |
|
| 95 |
+
| crop |  |  |
|
| 96 |
+
| monochrome, comic, 31.1 dB |  |  |
|
| 97 |
+
| crop |  |  |
|
| 98 |
+
|
| 99 |
+
On the monochrome crop the VAE keeps the screentone dots on the shorts; the head renders them as
|
| 100 |
+
flat grey (see Known limits).
|
| 101 |
+
|
| 102 |
+
## Held-out reads
|
| 103 |
+
|
| 104 |
Held-out reads during training (dB, colour / mono):
|
| 105 |
|
| 106 |
| decode point | colour | mono |
|
|
|
|
| 121 |
|
| 122 |
## Use
|
| 123 |
|
| 124 |
+
- ComfyUI: the [ComfyUI-Anima-PixelDecoder](https://github.com/sorryhyun/ComfyUI-Anima-PixelDecoder)
|
| 125 |
+
node pack (ComfyUI Manager: "Anima Pixel Decoder"). Put this file in `models/pixel_heads/`,
|
| 126 |
add **Anima Pixel Head Tap** on the MODEL before the KSampler (stock or Spectrum) and decode
|
| 127 |
the sampler's LATENT with **Anima Pixel Decode** instead of VAE Decode. Run the DiT on the
|
| 128 |
fp32 stream (Anima Adapter Loader, `stream = fp32`).
|