stability_sdk 0.1.0 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: dac029de10ccef05e11d7fee6a581f14fec50252e1cf7790ce2e849e5061489c
-  data.tar.gz: 66f8803ff804dae6971cdf3b8473225012255e87d67f00484583d65ecd8afb6a
+  metadata.gz: 2b707c869873f4d9ff68de1d281a70b498fb9d0509f83e6e3b3550e8f1d86bff
+  data.tar.gz: 04454b501ae8d8d4ed62140812b2f0482e9178fb54a7d4fd42b1e7bf2817f1fb
 SHA512:
-  metadata.gz: ad988347184eafe784ab654b0022f1742cf56e51c9e2d4d88b3507f97162c6aaa7a165ad8ee5a6d3c34c1a9427fc860a54679192235c154bfea3008765dd8f89
-  data.tar.gz: d4522e6b2644f0ef309dbfb862a0667ed80c77e93c30bc575764997aa511d73789b3f270641d44c83cc2fdc017b7a0d5562e7de79ed7bebf86ff894019ad2642
+  metadata.gz: efef7f4abc46b1664a4545514598c888472e7828a67d9ca93f5597c6cf3f1709cbc58bd1508936696626985689762b8d9100f773b1175f1a38fde8e507961df0
+  data.tar.gz: ca8e9fbc0ab70bb2f05b5171985c8159a4c423fe32c6a1903c42548366c1d1d17c40a172c85921e884291c64ef34e398bfda8f7e8e08af1632894504e292ede8

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    stability_sdk (0.1.0)
+    stability_sdk (0.2.0)
       grpc
       mime-types
 

data/README.md

@@ -36,6 +36,12 @@ This command saves an image like this:
 
 ![3749380973_A_night_in_winter__oil_on_canvas_landscape_painting__by_Vincent_van_Gogh](https://user-images.githubusercontent.com/25668/188884116-0b03494b-0b34-49de-bbbc-89fbc2f6029d.png)
 
+img2img example:
+
+```sh
+STABILITY_SDK_API_KEY=YOUR_API_KEY stability-client --init_image=/path/to/image.png --mask_image=/path/to/mask.png 'your prompt'
+```
+
 
 ```sh
 Usage: stability-client [options] YOUR_PROMPT_TEXT
@@ -52,6 +58,10 @@ Options:
         --no-store                   do not write out artifacts
     -n, --num_samples=VAL            number of samples to generate
     -e, --engine=VAL                 engine to use for inference. default `stable-diffusion-v1`
+    -i, --init_image=VAL             path to init image
+    -m, --mask_image=VAL             path to mask image
+        --start_schedule=VAL         start schedule for init image (must be greater than 0, 1 is full strength text prompt, no trace of image). default 1.0
+        --end_schedule=VAL           end schedule for init image. default 0.01
     -v, --verbose
 ```
 
@@ -67,6 +77,10 @@ client = StabilitySDK::Client.new(api_key: "YOUR_API_KEY")
 prompt = "your prompot here"
 options = {}
 
+# for the case of img2img:
+# options[:init_image] = "/path/to/image.png"
+# options[:mask_image] = "/path/to/mask.png"
+
 client.generate(prompt, options) do |answer|
   answer.artifacts.each do |artifact|
     if artifact.type == :ARTIFACT_IMAGE

data/exe/stability-client

@@ -27,14 +27,20 @@ opt.on("-p", "--prefix=VAL", "output prefixes for artifacts. default `generation
 opt.on("--no-store", "do not write out artifacts") {|v| options[:no_store] = v }
 opt.on("-n", "--num_samples=VAL", "number of samples to generate") {|v| options[:num_samples] = v }
 opt.on("-e", "--engine=VAL", "engine to use for inference. default `stable-diffusion-v1`") {|v| options[:engine_id] = v }
+opt.on("-i", "--init_image=VAL", "path to init image") {|v| options[:init_image] = v }
+opt.on("-m", "--mask_image=VAL", "path to mask image") {|v| options[:mask_image] = v }
+opt.on("--start_schedule=VAL", "start schedule for init image (must be greater than 0, 1 is full strength text prompt, no trace of image). default 1.0") {|v| options[:start_schedule] = v }
+opt.on("--end_schedule=VAL", "end schedule for init image. default 0.01") {|v| options[:end_schedule] = v }
 opt.on("-v", "--verbose") { logger.level = Logger::DEBUG }
 opt.parse!(ARGV)
 
 prompt = ARGV.join(" ")
-raise StabilitySDK::InsufficientParameter, "prompt is required" if prompt.nil? || prompt == ""
+raise StabilitySDK::InsufficientParameter, "prompt and/or init_image is required" if (prompt.nil? || prompt == "") && !options.has_key?(:init_image)
 
 options[:api_key] = ENV["STABILITY_SDK_API_KEY"] if ENV["STABILITY_SDK_API_KEY"]
-raise StabilitySDK::InsufficientParameter, "api key is required" if !options.has_key?(:api_key) || options[:api_key] == ""
+raise StabilitySDK::InsufficientParameter, "api key is required" if !options.has_key?(:api_key)
+
+raise StabilitySDK::InsufficientParameter, "init_image is required if mask_image is provided" if options.has_key?(:mask_image) && !options.has_key?(:init_image)
 
 client = StabilitySDK::Client.new(api_key: options[:api_key])
 

data/lib/generation_pb.rb

@@ -64,9 +64,14 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
     add_message "gooseai.ConditionerParameters" do
       proto3_optional :vector_adjust_prior, :string, 1
     end
+    add_message "gooseai.ScheduleParameters" do
+      optional :start, :float, 1
+      optional :end, :float, 2
+    end
     add_message "gooseai.StepParameter" do
       optional :scaled_step, :float, 1
       proto3_optional :sampler, :message, 2, "gooseai.SamplerParameters"
+      proto3_optional :schedule, :message, 3, "gooseai.ScheduleParameters"
     end
     add_message "gooseai.TransformType" do
       oneof :type do
@@ -125,6 +130,7 @@ Google::Protobuf::DescriptorPool.generated_pool.build do
       value :ARTIFACT_TOKENS, 4
       value :ARTIFACT_EMBEDDING, 5
       value :ARTIFACT_CLASSIFICATIONS, 6
+      value :ARTIFACT_MASK, 7
     end
     add_enum "gooseai.DiffusionSampler" do
       value :SAMPLER_DDIM, 0
@@ -166,6 +172,7 @@ module Gooseai
   Answer = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.Answer").msgclass
   SamplerParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.SamplerParameters").msgclass
   ConditionerParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ConditionerParameters").msgclass
+  ScheduleParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ScheduleParameters").msgclass
   StepParameter = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.StepParameter").msgclass
   TransformType = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.TransformType").msgclass
   ImageParameters = ::Google::Protobuf::DescriptorPool.generated_pool.lookup("gooseai.ImageParameters").msgclass

data/lib/stability_sdk/client.rb

@@ -8,9 +8,11 @@ module StabilitySDK
     DEFAULT_IMAGE_HEIGHT = 512
     DEFAULT_SAMPLE_SIZE = 1
     DEFAULT_STEPS = 50
-    DEFAULT_ENGINE_ID = "stable-diffusion-v1"
+    DEFAULT_ENGINE_ID = "stable-diffusion-v1-5"
     DEFAULT_CFG_SCALE = 7.0
     DEFAULT_SAMPLER_ALGORITHM = Gooseai::DiffusionSampler::SAMPLER_K_LMS
+    DEFAULT_START_SCHEDULE = 1.0
+    DEFAULT_END_SCHEDULE = 0.01
 
     sampler_algorithms = {
       "ddim": Gooseai::DiffusionSampler::SAMPLER_DDIM,
@@ -33,21 +35,8 @@ module StabilitySDK
     end
 
     def generate(prompt, options, &block)
-      image_param = image_param(options)
-      req = Gooseai::Request.new(
-        prompt: [Gooseai::Prompt.new(text: prompt)],
-        engine_id: options[:engine_id] || DEFAULT_ENGINE_ID,
-        image: image_param
-      )
-
-      @stub.generate(req).each do |answer|
-        block.call(answer)
-      end
-    end
-
-    def image_param(options={})
       width = options.has_key?(:width) ? options[:width].to_i : DEFAULT_IMAGE_WIDTH
-      height = options.has_key?(:height) ? options[:height] : DEFAULT_IMAGE_HEIGHT
+      height = options.has_key?(:height) ? options[:height].to_i : DEFAULT_IMAGE_HEIGHT
       samples = options.has_key?(:num_samples) ? [:num_samples].to_i : DEFAULT_SAMPLE_SIZE
       steps = options.has_key?(:steps) ? options[:steps].to_i : DEFAULT_STEPS
       seed = options.has_key?(:seed) ? [options[:seed]] : [rand(4294967295)]
@@ -61,7 +50,28 @@ module StabilitySDK
         ),
       )]
 
-      return Gooseai::ImageParameters.new(
+      prompt_param = []
+      if prompt != ""
+        prompt_param << Gooseai::Prompt.new(text: prompt)
+      end
+      if options.has_key?(:init_image)
+        prompt_param << init_image_to_prompt(options[:init_image])
+        parameters = [Gooseai::StepParameter.new(
+          scaled_step: 0,
+          sampler: Gooseai::SamplerParameters.new(
+            cfg_scale: options.has_key?(:cfg_scale) ? options[:cfg_scale].to_f : DEFAULT_CFG_SCALE,
+          ),
+          schedule: Gooseai::ScheduleParameters.new(
+            start: options.has_key?(:start_schedule) ? options[:start_schedule].to_f : DEFAULT_START_SCHEDULE,
+            end: options.has_key?(:end_schedule) ? options[:end_schedule].to_f : DEFAULT_END_SCHEDULE,
+          ),
+        )]
+      end
+      if options.has_key?(:mask_image)
+        prompt_param << mask_image_to_prompt(options[:mask_image])
+      end
+
+      image_param = Gooseai::ImageParameters.new(
         width: width,
         height: height,
         samples: samples,
@@ -70,6 +80,39 @@ module StabilitySDK
         transform: transform,
         parameters: parameters,
       )
+
+      req = Gooseai::Request.new(
+        prompt: prompt_param,
+        engine_id: options[:engine_id] || DEFAULT_ENGINE_ID,
+        image: image_param
+      )
+
+      @stub.generate(req).each do |answer|
+        block.call(answer)
+      end
+    end
+
+    def init_image_to_prompt(path)
+      bin = IO.binread(path)
+      return Gooseai::Prompt.new(
+        artifact: Gooseai::Artifact.new(
+          type: Gooseai::ArtifactType::ARTIFACT_IMAGE,
+          binary: bin,
+        ),
+        parameters: Gooseai::PromptParameters.new(
+          init: true
+        ),
+      )
+    end
+
+    def mask_image_to_prompt(path)
+      bin = IO.binread(path)
+      return Gooseai::Prompt.new(
+        artifact: Gooseai::Artifact.new(
+          type: Gooseai::ArtifactType::ARTIFACT_MASK,
+          binary: bin,
+        ),
+      )
     end
   end
 end

data/lib/stability_sdk/version.rb

@@ -1,3 +1,3 @@
 module StabilitySDK
-  VERSION = "0.1.0"
+  VERSION = "0.2.0"
 end

data/proto/generation.proto

@@ -19,6 +19,7 @@ enum ArtifactType {
   ARTIFACT_TOKENS = 4;
   ARTIFACT_EMBEDDING = 5;
   ARTIFACT_CLASSIFICATIONS = 6;
+  ARTIFACT_MASK = 7;
 }
 
 message Token {
@@ -106,9 +107,15 @@ enum Upscaler {
   UPSCALER_ESRGAN = 2;
 }
 
+message ScheduleParameters {
+  float start = 1;
+  float end = 2;
+}
+
 message StepParameter {
   float scaled_step = 1;
   optional SamplerParameters sampler = 2;
+  optional ScheduleParameters schedule = 3;
 }
 
 message TransformType {

data/stability_sdk.gemspec

@@ -7,7 +7,7 @@ Gem::Specification.new do |spec|
   spec.email         = ["cou929@gmail.com"]
 
   spec.summary       = "Ruby client for interacting with stability.ai APIs (e.g. stable diffusion inference)"
-  spec.description   = "Ruby client of https://github.com/Stability-AI/stability-sdk"
+  spec.description   = "Interacting with stability.ai APIs (e.g. stable diffusion inference). Ruby client of https://github.com/Stability-AI/stability-sdk ."
   spec.homepage      = "https://github.com/cou929/stability-sdk-ruby"
   spec.required_ruby_version = Gem::Requirement.new(">= 2.3.0")
 

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: stability_sdk
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.2.0
 platform: ruby
 authors:
 - Kosei Moriyama
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2022-09-07 00:00:00.000000000 Z
+date: 2022-09-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: grpc
@@ -52,7 +52,8 @@ dependencies:
     - - ">="
       - !ruby/object:Gem::Version
         version: '0'
-description: Ruby client of https://github.com/Stability-AI/stability-sdk
+description: Interacting with stability.ai APIs (e.g. stable diffusion inference).
+  Ruby client of https://github.com/Stability-AI/stability-sdk .
 email:
 - cou929@gmail.com
 executables: