(A1111) Img2Img inpaint (#147)

* Draw in paint screen

* InPaint parameters form

* Implement proper cropping

* Fix inpaint overlay

* InPaint prototype ready

* Translations

* Handle mode

Author: ShiftHackZ

data/src/main/java/com/shifthackz/aisdv1/data/mappers/ImageToImagePayloadMappers.kt

@@ -11,6 +11,13 @@ import java.util.Date
 fun ImageToImagePayload.mapToRequest(): ImageToImageRequest = with(this) {
     ImageToImageRequest(
         initImages = listOf(base64Image),
+        includeInitImages = true,
+        mask = base64MaskImage.takeIf(String::isNotBlank),
+        inPaintingMaskInvert = inPaintingMaskInvert,
+        inPaintFullResPadding = inPaintFullResPadding,
+        inPaintingFill = inPaintingFill,
+        inPaintFullRes = inPaintFullRes,
+        maskBlur = maskBlur,
         denoisingStrength = denoisingStrength,
         prompt = prompt,
         negativePrompt = negativePrompt,

dependencies.gradle

@@ -28,6 +28,8 @@ ext {
     cryptoVersion = '1.0.0'
     onnxruntimeVersion = '1.16.3'
     catppuccinVersion = '0.1.1'
+    composeGesturesVersion = '3.1'
+    composeEasyCropVersion = '0.1.1'
 
     testJunitVersion = '4.13.2'
 
@@ -92,6 +94,8 @@ ext {
             catppuccinLegacy      : "com.github.ShiftHackZ.Catppuccin-Android-Library:palette-legacy:$catppuccinVersion",
             catppuccinCompose     : "com.github.ShiftHackZ.Catppuccin-Android-Library:compose:$catppuccinVersion",
             catppuccinSplashscreen: "com.github.ShiftHackZ.Catppuccin-Android-Library:splashscreen:$catppuccinVersion",
+            composeGestures       : "com.github.SmartToolFactory:Compose-Extended-Gestures:$composeGesturesVersion",
+            composeEasyCrop       : "io.github.mr0xf00:easycrop:$composeEasyCropVersion",
     ]
     log = [
             timber: "com.jakewharton.timber:timber:$timberVersion",

domain/src/main/java/com/shifthackz/aisdv1/domain/entity/ImageToImagePayload.kt

@@ -2,6 +2,7 @@ package com.shifthackz.aisdv1.domain.entity
 
 data class ImageToImagePayload(
     val base64Image: String,
+    val base64MaskImage: String,
     val denoisingStrength: Float,
     val prompt: String,
     val negativePrompt: String,
@@ -16,4 +17,9 @@ data class ImageToImagePayload(
     val sampler: String,
     val nsfw: Boolean,
     val batchCount: Int,
+    val inPaintingMaskInvert: Int,
+    val inPaintFullResPadding: Int,
+    val inPaintingFill: Int,
+    val inPaintFullRes: Boolean,
+    val maskBlur: Int,
 )

network/src/main/java/com/shifthackz/aisdv1/network/request/ImageToImageRequest.kt

@@ -5,6 +5,20 @@ import com.google.gson.annotations.SerializedName
 data class ImageToImageRequest(
     @SerializedName("init_images")
     val initImages: List<String>,
+    @SerializedName("include_init_images")
+    val includeInitImages: Boolean,
+    @SerializedName("mask")
+    val mask: String?,
+    @SerializedName("inpainting_mask_invert")
+    val inPaintingMaskInvert: Int?,
+    @SerializedName("inpaint_full_res_padding")
+    val inPaintFullResPadding: Int?,
+    @SerializedName("inpainting_fill")
+    val inPaintingFill: Int?,
+    @SerializedName("inpaint_full_res")
+    val inPaintFullRes: Boolean?,
+    @SerializedName("mask_blur")
+    val maskBlur: Int?,
     @SerializedName("denoising_strength")
     val denoisingStrength: Float,
     @SerializedName("prompt")

presentation/build.gradle

@@ -43,5 +43,7 @@ dependencies {
     implementation ui.catppuccinCompose
     implementation ui.dayNightSwitch
     implementation ui.catppuccinSplashscreen
+    implementation ui.composeGestures
+    implementation ui.composeEasyCrop
     implementation "androidx.exifinterface:exifinterface:1.3.6"
 }

presentation/src/main/java/com/shifthackz/aisdv1/presentation/core/GenerationMviIntent.kt

@@ -1,5 +1,6 @@
 package com.shifthackz.aisdv1.presentation.core
 
+import android.graphics.Bitmap
 import com.shifthackz.aisdv1.domain.entity.AiGenerationResult
 import com.shifthackz.aisdv1.domain.entity.OpenAiModel
 import com.shifthackz.aisdv1.domain.entity.OpenAiQuality
@@ -86,13 +87,17 @@ sealed interface GenerationMviIntent : MviIntent {
 
 sealed interface ImageToImageIntent : GenerationMviIntent {
 
+    data object InPaint : ImageToImageIntent
+
     data object FetchRandomPhoto : ImageToImageIntent
 
     data object ClearImageInput : ImageToImageIntent
 
     data class UpdateDenoisingStrength(val value: Float) : ImageToImageIntent
 
-    data class UpdateImage(val result: PickedResult) : ImageToImageIntent
+    data class UpdateImage(val bitmap: Bitmap) : ImageToImageIntent
+
+    data class CropImage(val result: PickedResult) : ImageToImageIntent
 
     enum class Pick : ImageToImageIntent {
         Camera, Gallery

presentation/src/main/java/com/shifthackz/aisdv1/presentation/di/UiUtilsModule.kt

@@ -7,6 +7,7 @@ import com.shifthackz.aisdv1.presentation.screen.debug.DebugMenuAccessor
 import com.shifthackz.aisdv1.presentation.screen.gallery.detail.GalleryDetailBitmapExporter
 import com.shifthackz.aisdv1.presentation.screen.gallery.detail.GalleryDetailSharing
 import com.shifthackz.aisdv1.presentation.screen.gallery.list.GalleryExporter
+import com.shifthackz.aisdv1.presentation.screen.inpaint.InPaintStateProducer
 import org.koin.android.ext.koin.androidContext
 import org.koin.core.module.dsl.factoryOf
 import org.koin.core.module.dsl.singleOf
@@ -20,4 +21,5 @@ internal val uiUtilsModule = module {
     factoryOf(::GalleryDetailSharing)
     singleOf(::GenerationFormUpdateEvent)
     singleOf(::DebugMenuAccessor)
+    singleOf(::InPaintStateProducer)
 }

presentation/src/main/java/com/shifthackz/aisdv1/presentation/di/ViewModelModule.kt

@@ -11,6 +11,7 @@ import com.shifthackz.aisdv1.presentation.screen.gallery.detail.GalleryDetailVie
 import com.shifthackz.aisdv1.presentation.screen.gallery.list.GalleryViewModel
 import com.shifthackz.aisdv1.presentation.screen.home.HomeNavigationViewModel
 import com.shifthackz.aisdv1.presentation.screen.img2img.ImageToImageViewModel
+import com.shifthackz.aisdv1.presentation.screen.inpaint.InPaintViewModel
 import com.shifthackz.aisdv1.presentation.screen.loader.ConfigurationLoaderViewModel
 import com.shifthackz.aisdv1.presentation.screen.settings.SettingsViewModel
 import com.shifthackz.aisdv1.presentation.screen.setup.ServerSetupLaunchSource
@@ -40,6 +41,7 @@ val viewModelModule = module {
     viewModelOf(::ExtrasViewModel)
     viewModelOf(::EmbeddingViewModel)
     viewModelOf(::EditTagViewModel)
+    viewModelOf(::InPaintViewModel)
 
     viewModel { parameters ->
         val launchSource = ServerSetupLaunchSource.fromKey(parameters.get())

presentation/src/main/java/com/shifthackz/aisdv1/presentation/modal/ModalRenderer.kt

@@ -16,6 +16,8 @@ import com.shifthackz.aisdv1.core.model.UiText
 import com.shifthackz.aisdv1.core.model.asUiText
 import com.shifthackz.aisdv1.presentation.R
 import com.shifthackz.aisdv1.presentation.core.GenerationMviIntent
+import com.shifthackz.aisdv1.presentation.core.ImageToImageIntent
+import com.shifthackz.aisdv1.presentation.modal.crop.CropImageModal
 import com.shifthackz.aisdv1.presentation.modal.embedding.EmbeddingScreen
 import com.shifthackz.aisdv1.presentation.modal.extras.ExtrasScreen
 import com.shifthackz.aisdv1.presentation.modal.history.InputHistoryScreen
@@ -24,6 +26,7 @@ import com.shifthackz.aisdv1.presentation.modal.tag.EditTagDialog
 import com.shifthackz.aisdv1.presentation.model.Modal
 import com.shifthackz.aisdv1.presentation.screen.gallery.detail.GalleryDetailIntent
 import com.shifthackz.aisdv1.presentation.screen.gallery.list.GalleryIntent
+import com.shifthackz.aisdv1.presentation.screen.inpaint.InPaintIntent
 import com.shifthackz.aisdv1.presentation.screen.settings.SettingsIntent
 import com.shifthackz.aisdv1.presentation.screen.setup.ServerSetupIntent
 import com.shifthackz.aisdv1.presentation.widget.dialog.DecisionInteractiveDialog
@@ -46,6 +49,7 @@ fun ModalRenderer(
         processIntent(GenerationMviIntent.SetModal(Modal.None))
         processIntent(GalleryIntent.DismissDialog)
         processIntent(GalleryDetailIntent.DismissDialog)
+        processIntent(InPaintIntent.ScreenModal.Dismiss)
     }
     when (screenModal) {
         Modal.None -> Unit
@@ -223,5 +227,20 @@ fun ModalRenderer(
             onConfirmAction = { processIntent(ServerSetupIntent.LocalModel.DeleteConfirm(screenModal.model)) },
             onDismissRequest = dismiss,
         )
+
+        Modal.ClearInPaintConfirm -> DecisionInteractiveDialog(
+            title = R.string.interaction_in_paint_clear_title.asUiText(),
+            text = R.string.interaction_in_paint_clear_title.asUiText(),
+            confirmActionResId = R.string.yes,
+            dismissActionResId = R.string.no,
+            onConfirmAction = { processIntent(InPaintIntent.Action.Clear) },
+            onDismissRequest = dismiss,
+        )
+
+        is Modal.Image.Crop -> CropImageModal(
+            bitmap = screenModal.bitmap,
+            onDismissRequest = dismiss,
+            onResult = { processIntent(ImageToImageIntent.UpdateImage(it)) }
+        )
     }
 }

presentation/src/main/java/com/shifthackz/aisdv1/presentation/modal/crop/CropImageModal.kt

@@ -0,0 +1,73 @@
+package com.shifthackz.aisdv1.presentation.modal.crop
+
+import android.graphics.Bitmap
+import androidx.compose.material3.MaterialTheme
+import androidx.compose.runtime.Composable
+import androidx.compose.runtime.LaunchedEffect
+import androidx.compose.ui.graphics.asAndroidBitmap
+import androidx.compose.ui.graphics.asImageBitmap
+import androidx.compose.ui.platform.LocalContext
+import com.mr0xf00.easycrop.AspectRatio
+import com.mr0xf00.easycrop.CropError
+import com.mr0xf00.easycrop.CropResult
+import com.mr0xf00.easycrop.CropperStyle
+import com.mr0xf00.easycrop.CropperStyleGuidelines
+import com.mr0xf00.easycrop.RectCropShape
+import com.mr0xf00.easycrop.crop
+import com.mr0xf00.easycrop.rememberImageCropper
+import com.mr0xf00.easycrop.ui.ImageCropperDialog
+import com.shifthackz.aisdv1.core.extensions.showToast
+
+@Composable
+fun CropImageModal(
+    bitmap: Bitmap,
+    onResult: (Bitmap) -> Unit = {},
+    onDismissRequest: () -> Unit = {},
+) {
+    val imageCropper = rememberImageCropper()
+    val state = imageCropper.cropState
+    state?.let {
+        LaunchedEffect(Unit) {
+            it.region = when {
+                it.region.height > it.region.width -> it.region.copy(
+                    bottom = it.region.width,
+                )
+
+                it.region.width > it.region.height -> it.region.copy(
+                    right = it.region.height
+                )
+
+                else -> it.region
+            }
+            it.aspectLock = true
+        }
+        ImageCropperDialog(
+            state = it,
+            style = CropperStyle(
+                backgroundColor = MaterialTheme.colorScheme.background,
+                overlay = MaterialTheme.colorScheme.surface,
+                guidelines = CropperStyleGuidelines(),
+                shapes = listOf(RectCropShape),
+                aspects = listOf(AspectRatio(1, 1)),
+            ),
+        )
+    }
+    val context = LocalContext.current
+    LaunchedEffect(Unit) {
+        when (val result = imageCropper.crop(bmp = bitmap.asImageBitmap())) {
+            is CropResult.Success -> result.bitmap.asAndroidBitmap().let(onResult::invoke)
+
+            CropError.LoadingError -> {
+                context.showToast("Loading error")
+                onDismissRequest()
+            }
+
+            CropError.SavingError -> {
+                context.showToast("Saving error")
+                onDismissRequest()
+            }
+
+            CropResult.Cancelled -> onDismissRequest()
+        }
+    }
+}