You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 3bd5e07
Browse filesBrowse the repository at this point in the historyBrowse files
@@ -10,34 +10,36 @@ an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express o
10
10
specific language governing permissions and limitations under the License.
11
11
-->
12
12
13
-
# Weighting prompts
13
+
# Prompt weighting
14
14
15
15
[[open-in-colab]]
16
16
17
-
Text-guided diffusion models generate images based on a given text prompt. The text prompt
18
-
can include multiple concepts that the model should generate and it's often desirable to weight
19
-
certain parts of the prompt more or less.
17
+
Prompt weighting provides a way to emphasize or de-emphasize certain parts of a prompt, allowing for more control over the generated image. A prompt can include several concepts, which gets turned into contextualized text embeddings. The embeddings are used by the model to condition its cross-attention layers to generate an image (read the Stable Diffusion [blog post](https://huggingface.co/blog/stable_diffusion) to learn more about how it works).
20
18
21
-
Diffusion models work by conditioning the cross attention layers of the diffusion model with contextualized text embeddings (see the [Stable Diffusion Guide for more information](../stable-diffusion)).
22
-
Thus a simple way to emphasize (or de-emphasize) certain parts of the prompt is by increasing or reducing the scale of the text embedding vector that corresponds to the relevant part of the prompt.
23
-
This is called "prompt-weighting" and has been a highly demanded feature by the community (see issue [here](https://github.com/huggingface/diffusers/issues/2431)).
19
+
Prompt weighting works by increasing or decreasing the scale of the text embedding vector that corresponds to its concept in the prompt because you may not necessarily want the model to focus on all concepts equally. The easiest way to prepare the prompt-weighted embeddings is to use [Compel](https://github.com/damian0815/compel), a text prompt-weighting and blending library. Once you have the prompt-weighted embeddings, you can pass them to any pipeline that has a [`prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/text2img#diffusers.StableDiffusionPipeline.__call__.prompt_embeds) (and optionally [`negative_prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/text2img#diffusers.StableDiffusionPipeline.__call__.negative_prompt_embeds)) parameter, such as [`StableDiffusionPipeline`], [`StableDiffusionControlNetPipeline`], and [`StableDiffusionXLPipeline`].
24
20
25
-
## How to do prompt-weighting in Diffusers
21
+
<Tip>
26
22
27
-
We believe the role of `diffusers` is to be a toolbox that provides essential features that enable other projects, such as [InvokeAI](https://github.com/invoke-ai/InvokeAI) or [diffuzers](https://github.com/abhishekkrthakur/diffuzers), to build powerful UIs. In order to support arbitrary methods to manipulate prompts, `diffusers` exposes a [`prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/text2img#diffusers.StableDiffusionPipeline.__call__.prompt_embeds) function argument and an optional [`negative_prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/text2img#diffusers.StableDiffusionPipeline.__call__.negative_prompt_embeds) function argument to many pipelines such as [`StableDiffusionPipeline`], [`StableDiffusionControlNetPipeline`], [`StableDiffusionXLPipeline`], allowing to directly pass the "prompt-weighted"/scaled text embeddings to the pipeline.
23
+
If your favorite pipeline doesn't have a `prompt_embeds` parameter, please open an [issue](https://github.com/huggingface/diffusers/issues/new/choose) so we can add it!
28
24
29
-
The [compel library](https://github.com/damian0815/compel) provides an easy way to emphasize or de-emphasize portions of the prompt for you. We strongly recommend it instead of preparing the embeddings yourself.
25
+
</Tip>
30
26
31
-
Let's look at a simple example. Imagine you want to generate an image of `"a red cat playing with a ball"` as
32
-
follows:
27
+
This guide will show you how to weight and blend your prompts with Compel in 🤗 Diffusers.
33
28
29
+
Before you begin, make sure you have the latest version of Compel installed:
34
30
35
-
### StableDiffusionPipeline
31
+
```py
32
+
# uncomment to install in Colab
33
+
#!pip install compel --upgrade
34
+
```
35
+
36
+
For this guide, let's generate an image with the prompt `"a red cat playing with a ball"` using the [`StableDiffusionPipeline`]:
36
37
37
38
```py
38
39
from diffusers import StableDiffusionPipeline, UniPCMultistepScheduler
You'll notice there is no "ball" in the image! Let's use compel to upweight the concept of "ball" in the prompt. Create a [`Compel`](https://github.com/damian0815/compel/blob/main/doc/compel.md#compel-objects) object, and pass it a tokenizer and text encoder:
54
60
55
-
As you can see, there is no "ball" in the image. Let's emphasize this part!
compel uses `+` or `-` to increase or decrease the weight of a word in the prompt. To increase the weight of "ball":
68
+
69
+
<Tip>
70
+
71
+
`+` corresponds to the value `1.1`, `++` corresponds to `1.1^2`, and so on. Similarly, `-` corresponds to `0.9` and `--` corresponds to `0.9^2`. Feel free to experiment with adding more `+` or `-` in your prompt!
72
+
73
+
</Tip>
58
74
59
75
```py
60
-
pip install compel --upgrade
76
+
prompt ="a red cat playing with a ball++"
61
77
```
62
78
63
-
and then create a `Compel` object:
79
+
Pass the prompt to `compel_proc` to create the new prompt embeddings which are passed to the pipeline:
You can also create a weighted *blend* of prompts by adding `.blend()` to a list of prompts and passing it some weights. Your blend may not always produce the result you expect because it breaks some assumptions about how the text encoder functions, so just have fun and experiment with it!
Similarly, we de-emphasize parts of the sentence by using the `--` suffix for words, feel free to give it
97
-
a try!
143
+
A conjunction diffuses each prompt independently and concatenates their results by their weighted sum. Add `.and()` to the end of a list of prompts to create a conjunction:
144
+
145
+
```py
146
+
prompt_embeds = compel_proc('("a red cat, playing with a, ball").and()')
[Textual inversion](../training/text_inversion) is a technique for learning a specific concept from some images which you can use to generate new images conditioned on that concept.
101
160
102
-
Compel 1.1.6 adds a utility class to simplify using textual inversions. Instantiate a `DiffusersTextualInversionManager` and pass it to Compel init:
161
+
Create a pipeline and use the [`~loaders.TextualInversionLoaderMixin.load_textual_inversion`] function to load the textual inversion embeddings (feel free to browse the [Stable Diffusion Conceptualizer](https://huggingface.co/spaces/sd-concepts-library/stable-diffusion-conceptualizer) for 100+ trained concepts):
103
162
163
+
```py
164
+
import torch
165
+
from diffusers import StableDiffusionPipeline
166
+
from compel import Compel, DiffusersTextualInversionManager
Compel provides a `DiffusersTextualInversionManager` class to simplify prompt weighting with textual inversion. Instantiate `DiffusersTextualInversionManager` and pass it to the `Compel` class:
[DreamBooth](../training/dreambooth) is a technique for generating contextualized images of a subject given just a few images of the subject to train on. It is similar to textual inversion, but DreamBooth trains the full model whereas textual inversion only fine-tunes the text embeddings. This means you should use [`~DiffusionPipeline.from_pretrained`] to load the DreamBooth model (feel free to browse the [Stable Diffusion Dreambooth Concepts Library](https://huggingface.co/sd-dreambooth-library) for 100+ trained models):
198
+
199
+
```py
200
+
import torch
201
+
from diffusers import DiffusionPipeline, UniPCMultistepScheduler
Create a `Compel` class with a tokenizer and text encoder, and pass your prompt to it. Depending on the model you use, you'll need to incorporate the model's unique identifier into your prompt. For example, the `dndcoverart-v1` model uses the identifier `dndcoverart`:
For StableDiffusionXL we need to not only pass `prompt_embeds` (and optionally `negative_prompt_embeds`), but also [`pooled_prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLInpaintPipeline.__call__.pooled_prompt_embeds) and optionally [`negative_pooled_prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLInpaintPipeline.__call__.negative_pooled_prompt_embeds).
118
-
In addition, [`StableDiffusionXLPipeline`] has two tokenizers and two text encoders which both need to be used to weight the prompt.
119
-
Luckily, [`compel`](https://github.com/damian0815/compel) takes care of SDXL's special needs - all we have to do is to pass both tokenizers and text encoders to the `Compel` class.
221
+
## Stable Diffusion XL
120
222
223
+
Stable Diffusion XL (SDXL) has two tokenizers and text encoders so it's usage is a bit different. To address this, you should pass both tokenizers and encoders to the `Compel` class:
121
224
122
225
```py
123
226
from compel import Compel, ReturnedEmbeddingsType
@@ -138,32 +241,25 @@ compel = Compel(
138
241
)
139
242
```
140
243
141
-
Let's try our example from above again. We use the same seed for both prompts and upweight ball by a factor of 1.5 for the first
142
-
prompt and downweight ball by 40% for the second prompt.
244
+
This time, let's upweight "ball" by a factor of 1.5 for the first prompt, and downweight "ball" by 0.6 for the second prompt. The [`StableDiffusionXLPipeline`] also requires [`pooled_prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLInpaintPipeline.__call__.pooled_prompt_embeds) (and optionally [`negative_pooled_prompt_embeds`](https://huggingface.co/docs/diffusers/en/api/pipelines/stable_diffusion/stable_diffusion_xl#diffusers.StableDiffusionXLInpaintPipeline.__call__.negative_pooled_prompt_embeds)) so you should pass those to the pipeline along with the conditioning tensors:
143
245
144
246
```py
145
-
#upweight "ball"
247
+
#apply weights
146
248
prompt = ["a red cat playing with a (ball)1.5", "a red cat playing with a (ball)0.6"]
147
249
conditioning, pooled = compel(prompt)
148
250
149
-
150
251
# generate image
151
252
generator = [torch.Generator().manual_seed(33) for _ inrange(len(prompt))]
<figcaption class="mt-2 text-center text-sm text-gray-500">a red cat playing with a (ball)0.6</figcaption>
263
+
<figcaption class="mt-2 text-center text-sm text-gray-500">"a red cat playing with a (ball)0.6"</figcaption>
165
264
</div>
166
-
</div>
167
-
168
-
We can see that the ball is almost completely gone on the right image while it's clearly visible on the left image.
169
-
For more information and more tricks you can use `compel` with, please have a look at the [compel docs](https://github.com/damian0815/compel/blob/main/doc/syntax.md) as well.
0 commit comments