2d aspect ratio

crates/desktop/artcraft/src/services/sora/commands/sora_image_remix_command.rs

@@ -42,7 +42,9 @@ const SORA_IMAGE_UPLOAD_TIMEOUT: Duration = Duration::from_millis(1000 * 30); //
 
 const SORA_IMAGE_REMIX_TIMEOUT: Duration = Duration::from_millis(1000 * 30); // 30 seconds
 
-#[derive(Deserialize)]
+const DEFAULT_ASPECT_RATIO : SoraAspectRatio = SoraAspectRatio::Square;
+
+#[derive(Deserialize, Debug)]
 pub struct SoraImageRemixCommand {
   /// Image media file; the engine or canvas snapshot (screenshot).
   pub snapshot_media_token: MediaFileToken,
@@ -57,6 +59,17 @@ pub struct SoraImageRemixCommand {
   pub maybe_additional_images: Option<Vec<MediaFileToken>>,
 
   pub maybe_number_of_samples: Option<u32>,
+  
+  /// Aspect ratio.
+  pub aspect_ratio: Option<SoraAspectRatio>,
+}
+
+#[derive(Deserialize, Debug, Copy, Clone)]
+#[serde(rename_all = "snake_case")]
+pub enum SoraAspectRatio {
+  Square,
+  Wide,
+  Tall,
 }
 
 #[derive(Serialize, Debug)]
@@ -74,7 +87,6 @@ pub enum SoraImageRemixErrorType {
   SoraIsHavingProblems,
 }
 
-
 #[tauri::command]
 pub async fn sora_image_remix_command(
   app: AppHandle,
@@ -84,8 +96,8 @@ pub async fn sora_image_remix_command(
   sora_task_queue: State<'_, SoraTaskQueue>,
 ) -> CommandResult<(), SoraImageRemixErrorType, ()> {
   
-  info!("image_generation_command called; scene media token: {:?}, additional images: {:?}", 
-    request.snapshot_media_token, request.maybe_additional_images);
+  info!("image_generation_command called; scene media token: {:?}, additional images: {:?}, full request: {:?}", 
+    &request.snapshot_media_token, &request.maybe_additional_images, &request);
 
   // TODO(bt,2025-04-24): Better error messages to caller
 
@@ -240,6 +252,14 @@ pub async fn generate_image(
     sora_media_tokens.push(response.id);
   }
 
+  let aspect_ratio = request.aspect_ratio.unwrap_or(DEFAULT_ASPECT_RATIO);
+  
+  let aspect_ratio = match aspect_ratio {
+    SoraAspectRatio::Square => ImageSize::Square,
+    SoraAspectRatio::Wide => ImageSize::Wide,
+    SoraAspectRatio::Tall => ImageSize::Tall,
+  };
+
   info!("Calling image generation...");
 
   // TODO(bt,2025-04-21): Download media tokens.
@@ -250,7 +270,7 @@ pub async fn generate_image(
       image_remix_with_session_auto_renew(ImageRemixAutoRenewRequest {
         prompt: request.prompt.to_string(),
         num_images: NumImages::One,
-        image_size: ImageSize::Square,
+        image_size: aspect_ratio,
         sora_media_tokens: sora_media_tokens.clone(),
         credentials: &creds,
         request_timeout: Some(SORA_IMAGE_REMIX_TIMEOUT),

frontend/libs/components/promptbox/src/lib/PromptBox2D.tsx

@@ -1,5 +1,4 @@
 import { useState, useRef, useEffect } from "react";
-import { invoke } from "@tauri-apps/api/core";
 import { useSignals } from "@preact/signals-react/runtime";
 import { toast } from "@storyteller/ui-toaster";
 import {
@@ -37,6 +36,7 @@ import {
   GetAppPreferences,
   SoraImageRemix,
   SoraImageRemixErrorType,
+  SoraImageRemixAspectRatio,
 } from "@storyteller/tauri-api";
 import { SoundRegistry } from "@storyteller/soundboard";
 
@@ -291,12 +291,15 @@ export const PromptBox2D = ({
     //console.log("Generate response:", generateResponse);
     //toast.success("Please wait while we process your image.");
 
+    const aspectRatio = getCurrentSoraRemixAspectRatio();
+
     const generateResponse = await SoraImageRemix({
       snapshot_media_token: snapshotMediaToken.data,
       disable_system_prompt: !useSystemPrompt,
       prompt: prompt,
       maybe_additional_images: referenceImages.map((image) => image.mediaToken),
       maybe_number_of_samples: 1,
+      aspect_ratio: aspectRatio,
     });
 
     console.log("generateResponse", generateResponse);
@@ -448,6 +451,19 @@ export const PromptBox2D = ({
     return iconElement.props.icon;
   };
 
+  const getCurrentSoraRemixAspectRatio = (): SoraImageRemixAspectRatio => {
+    const selected = aspectRatioList.find((item) => item.selected);
+    switch (selected?.label) {
+      case "3:2":
+        return SoraImageRemixAspectRatio.Wide;
+      case "2:3":
+        return SoraImageRemixAspectRatio.Tall;
+      case "1:1":
+      default:
+        return SoraImageRemixAspectRatio.Square;
+    }
+  };
+
   const handleKeyDown = (e: React.KeyboardEvent<HTMLTextAreaElement>) => {
     if (e.key === "Enter" && !e.shiftKey) {
       e.preventDefault();

frontend/libs/components/promptbox/src/lib/PromptBox3D.tsx

@@ -51,6 +51,7 @@ import {
   CommandSuccessStatus,
   GetAppPreferences,
   SoraImageRemix,
+  SoraImageRemixAspectRatio,
   SoraImageRemixErrorType,
 } from "@storyteller/tauri-api";
 
@@ -415,6 +416,8 @@ export const PromptBox3D = ({
 
         console.log("snapshotResult", snapshotResult);
 
+        const aspectRatio = getCurrentSoraRemixAspectRatio();
+
         const generateResponse = await SoraImageRemix({
           snapshot_media_token: snapshotResult.data!,
           disable_system_prompt: !useSystemPrompt,
@@ -423,6 +426,7 @@ export const PromptBox3D = ({
             (image) => image.mediaToken
           ),
           maybe_number_of_samples: 1,
+          aspect_ratio: aspectRatio,
         });
 
         console.log("generateResponse", generateResponse);
@@ -517,6 +521,19 @@ export const PromptBox3D = ({
     }
   };
 
+  const getCurrentSoraRemixAspectRatio = () : SoraImageRemixAspectRatio => {
+    switch (cameraAspectRatio.value) {
+      case CameraAspectRatio.HORIZONTAL_3_2:
+        return SoraImageRemixAspectRatio.Wide;
+      case CameraAspectRatio.VERTICAL_2_3:
+      case CameraAspectRatio.VERTICAL_9_16:
+        return SoraImageRemixAspectRatio.Tall;
+      case CameraAspectRatio.SQUARE_1_1:
+      default:
+        return SoraImageRemixAspectRatio.Square;
+    }
+  }
+
   const handleSaveFrame = async () => {
     if (!snapshotCurrentFrame) {
       return;

frontend/libs/tauri-api/src/lib/sora/SoraImageRemix.ts

@@ -17,6 +17,15 @@ export interface SoraImageRemixRequest {
 
   // Typically "1"
   maybe_number_of_samples: number,
+
+  // The aspect ratio of the output image
+  aspect_ratio?: SoraImageRemixAspectRatio,
+}
+
+export enum SoraImageRemixAspectRatio {
+  Square = "square",
+  Wide = "wide",
+  Tall = "tall",
 }
 
 export interface SoraImageRemixSuccess extends CommandResult {
@@ -43,13 +52,7 @@ export type SoraImageRemixResult = SoraImageRemixSuccess | SoraImageRemixError;
 export const SoraImageRemix = async (request: SoraImageRemixRequest) : Promise<SoraImageRemixResult> => {
   try {
     return await invoke("sora_image_remix_command", {
-        request: {
-        snapshot_media_token: request.snapshot_media_token,
-        disable_system_prompt: request.disable_system_prompt,
-        prompt: request.prompt,
-        maybe_additional_images: request.maybe_additional_images,
-        maybe_number_of_samples: request.maybe_number_of_samples,
-        },
+      request: request,
     }) as SoraImageRemixResult;
   } catch (error) {
     let maybeTypedError = error as CommandResult;