From 1e6bbf892855532b3a54642fde1292bf6f9c3401 Mon Sep 17 00:00:00 2001 From: Sean Sube Date: Sat, 10 Jun 2023 14:30:50 -0500 Subject: [PATCH] start updating user guide for v0.10 features --- docs/user-guide.md | 101 ++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 99 insertions(+), 2 deletions(-) diff --git a/docs/user-guide.md b/docs/user-guide.md index 96ce9403..6e63a67b 100644 --- a/docs/user-guide.md +++ b/docs/user-guide.md @@ -33,16 +33,34 @@ Please see [the server admin guide](server-admin.md) for details on how to confi - [LoRA tokens](#lora-tokens) - [Textual Inversion tokens](#textual-inversion-tokens) - [CLIP skip tokens](#clip-skip-tokens) - - [Long prompt weighting](#long-prompt-weighting) + - [Long prompt weighting syntax](#long-prompt-weighting-syntax) + - [Pipelines](#pipelines) + - [Stable Diffusion pipeline](#stable-diffusion-pipeline) + - [ControlNet pipeline](#controlnet-pipeline) + - [img2img pipeline](#img2img-pipeline) + - [Inpaint pipeline](#inpaint-pipeline) + - [Long prompt weighting pipeline](#long-prompt-weighting-pipeline) + - [Panorama pipeline](#panorama-pipeline) + - [Instruct pix2pix pipeline](#instruct-pix2pix-pipeline) + - [Txt2Img pipeline](#txt2img-pipeline) + - [Upscale pipeline](#upscale-pipeline) - [Tabs](#tabs) - [Txt2img tab](#txt2img-tab) - [Scheduler parameter](#scheduler-parameter) + - [Eta parameter](#eta-parameter) - [CFG parameter](#cfg-parameter) - [Steps parameter](#steps-parameter) - [Seed parameter](#seed-parameter) + - [Batch size parameter](#batch-size-parameter) + - [Tile size parameter](#tile-size-parameter) + - [Overlap parameter](#overlap-parameter) + - [UNet stride parameter](#unet-stride-parameter) + - [Tiled VAE parameter](#tiled-vae-parameter) - [Prompt parameter](#prompt-parameter) - [Negative prompt parameter](#negative-prompt-parameter) - [Width and height parameters](#width-and-height-parameters) + - [Highres parameters](#highres-parameters) + - [Upscale and correction parameters](#upscale-and-correction-parameters) - [Img2img tab](#img2img-tab) - [Img2img source image](#img2img-source-image) - [Strength parameter](#strength-parameter) @@ -58,6 +76,7 @@ Please see [the server admin guide](server-admin.md) for details on how to confi - [Denoise parameter](#denoise-parameter) - [Face correction and strength](#face-correction-and-strength) - [Blend tab](#blend-tab) + - [Models tab](#models-tab) - [Settings tab](#settings-tab) - [Image history setting](#image-history-setting) - [API server setting](#api-server-setting) @@ -283,7 +302,7 @@ You can skip the last layers of the CLIP text encoder using the `clip` token: This makes your prompt less specific and some models have been trained to work better with some amount of skipping. -### Long prompt weighting +### Long prompt weighting syntax You can emphasize or deemphasize certain parts of the prompt by using the long prompt weighting option. This adds some additional tokens: @@ -298,6 +317,46 @@ some additional tokens: _Note:_ The [token range syntax](#textual-inversion-tokens) currently does not work when long prompt weighting is enabled. +## Pipelines + +TODO + +### Stable Diffusion pipeline + +TODO + +### ControlNet pipeline + +TODO + +### img2img pipeline + +TODO + +### Inpaint pipeline + +TODO + +### Long prompt weighting pipeline + +TODO + +### Panorama pipeline + +TODO + +### Instruct pix2pix pipeline + +TODO + +### Txt2Img pipeline + +TODO + +### Upscale pipeline + +TODO + ## Tabs ### Txt2img tab @@ -312,6 +371,10 @@ This selects the scheduler algorithm used to resolve the latent noise into a coh See [the scheduler comparison](#scheduler-comparison) for more details. +#### Eta parameter + +TODO + #### CFG parameter Classifier free guidance. How strictly the model should follow the prompt. Anything from 5 to 15 usually works. More is @@ -343,6 +406,26 @@ You can use the same prompt and seed, while varying the steps and CFG, to produc Using -1 will generate a new seed on the server for each image. +#### Batch size parameter + +TODO + +#### Tile size parameter + +TODO + +#### Overlap parameter + +TODO + +#### UNet stride parameter + +TODO + +#### Tiled VAE parameter + +TODO + #### Prompt parameter The input text for your image, things that should be included. @@ -373,6 +456,14 @@ The opposite of [the prompt parameter](#prompt-parameter), things that should _n Controls the size of the output image, before upscaling. +#### Highres parameters + +TODO + +#### Upscale and correction parameters + +TODO: these allow you to use the upscale and face correction models with other pipelines + ### Img2img tab The img2img tab takes a source image along with the text prompt and produces a similar image. You can use the @@ -485,6 +576,8 @@ without running a diffusion pipeline at all. This can be faster and avoids makin Resize the output image before returning it to the client. +TODO: include other upscale models + Enabling this will run Real ESRGAN and requires an upscaling model. Check out [the Real ESRGAN Github](https://github.com/xinntao/Real-ESRGAN) for more details. @@ -519,6 +612,10 @@ in the mask and the second image will be used as the source for white pixels. Gr Upscaling and correction run after the images have been blended. +### Models tab + +TODO + ### Settings tab The settings tab provides access to some of the settings and allows you to reset the state of the other tabs