Spaces:
Running
on
Zero
Usage
Enter a prompt and click Generate. Read Civitai's guide on prompting to learn more.
Compel
Positive and negative prompts are embedded by Compel, enabling weighting and blending. See syntax features.
Embeddings
Textual inversion embeddings are installed for use in the Negative prompt.
- Bad Prompt:
<bad_prompt> - Negative Hand:
<negative_hand> - Fast Negative:
<fast_negative>- includes Negative Hand
- Bad Dream:
<bad_dream> - Unrealistic Dream:
<unrealistic_dream>- pair with Fast Negative and the Realistic Vision model
Arrays
Arrays allow you to generate different images from a single prompt. For example, a cute [[cat,corgi,koala]] will expand into 3 prompts. For this to work, you first have to increase Images. Note that it only works for the positive prompt. Inspired by Fooocus.
Autoincrement
If Autoincrement checked, the seed will be incremented for each image in range Images. When using arrays, you might want this disabled so the same seed is used.
Models
All use float16 (or bfloat16 if supported).
- fluently/fluently-v4
- linaqruf/anything-v3-1
- lykon/dreamshaper-8
- prompthero/openjourney-v4
- runwayml/stable-diffusion-v1-5
- sg161222/realistic_vision_v5.1
Schedulers
All are based on k_diffusion except DEIS and DPM++. Optionally, the Karras noise schedule can be used.
Advanced
DeepCache
DeepCache (Ma et al. 2023) caches UNet layers determined by Branch and reuses them every Interval steps. Leaving Branch on 0 caches lower layers, which provides a greater speedup. An Interval of 3 is the best balance between speed and quality; 1 means no cache.
T-GATE
T-GATE (Zhang et al. 2024) caches self and cross attention computations up to Step. Afterwards, attention is no longer computed and the cache is used, resulting in a noticeable speedup. Works well with DeepCache.
Tiny VAE
Enable madebyollin/taesd for almost instant latent decoding with a minor loss in detail. Useful for development and ideation.
Clip Skip
When enabled, the last CLIP layer is skipped. This can improve image quality and is commonly used with anime models.
Prompt Truncation
When enabled, prompts will be truncated to CLIP's limit of 77 tokens. By default this is disabled, so Compel will chunk prompts into segments rather than cutting them off.