Feat/ai refactor

schemas/audioasset.yaml

@@ -1,8 +1,15 @@
     AudioAsset:
-      description: >-
-        The AudioAsset is used to add sound effects and audio at specific intervals on
-        the timeline. The src must be a publicly accessible URL to an audio resource such 
-        as an mp3 file.
+      description: |
+        The AudioAsset adds audio to a Clip. The audio can be sourced from a URL
+        (`src`) or generated from a text prompt (`prompt`). Exactly one of `src` or
+        `prompt` must be provided.
+
+        - **Source URL:** set `src` to a publicly accessible audio URL (e.g. mp3).
+        - **Generated speech:** set `prompt` to the spoken text and `voice` to a voice
+          identifier (text-to-speech). Optionally set `language`/`newscaster`.
+        - **Generated music or SFX:** set `prompt` describing the sound; omit `voice`.
+        - Use `model` to choose the generator. The generated `src` is filled in
+          automatically.
       type: object
       properties:
         type:
@@ -13,11 +20,43 @@
         src:
           description: >-
             The audio source URL. The URL must be publicly accessible or include
-            credentials.
+            credentials. Provide either `src` or `prompt`, not both.
           type: string
           minLength: 1
           pattern: '\S'
           example: https://s3-ap-northeast-1.amazonaws.com/my-bucket/sound.mp3
+        prompt:
+          description: >-
+            A text prompt. When `voice` is set, the prompt is the spoken text
+            (text-to-speech). Without `voice`, the prompt describes generated music
+            or sound effects. The generated `src` is filled in automatically.
+          type: string
+          maxLength: 4000
+          example: Welcome to today's broadcast.
+        voice:
+          description: >-
+            Voice identifier for text-to-speech generation (e.g. `Matthew`, `Joanna`).
+            Only meaningful when `prompt` is set.
+          type: string
+          example: Matthew
+        language:
+          description: >-
+            Optional BCP-47 language code (e.g. `en-US`) for text-to-speech.
+            Only meaningful when `prompt` and `voice` are set.
+          type: string
+          example: en-US
+        newscaster:
+          description: >-
+            Set to `true` to use the voice's newscaster mode when supported.
+            Only meaningful when `prompt` and `voice` are set.
+          type: boolean
+          default: false
+        model:
+          description: >-
+            The generation model to use when `prompt` is set (e.g. `polly-neural`).
+            Defaults to the platform's preferred generator if omitted.
+          type: string
+          example: polly-neural
         trim:
           description: >-
             The start trim point of the audio clip, in seconds (defaults to 0). Audio
@@ -46,7 +85,7 @@
         speed:
           description: >-
             Adjust the playback speed of the audio clip between 0 (paused) and 10 (10x normal speed), where 1 is normal
-            speed (defaults to 1). Adjusting the speed will also adjust the duration of the clip and may require you to 
+            speed (defaults to 1). Adjusting the speed will also adjust the duration of the clip and may require you to
             adjust the Clip length. For example, if you set speed to 0.5, the clip will need to be 2x as long to play
             the entire audio (i.e. original length / 0.5). If you set speed to 2, the clip will need to be half as long
             to play the entire audio (i.e. original length / 2).
@@ -72,4 +111,3 @@
       additionalProperties: false
       required:
         - type
-        - src

schemas/imageasset.yaml

@@ -1,7 +1,12 @@
     ImageAsset:
-      description: >-
-        The ImageAsset is used to create video from images to compose an image. The src must be a
-        publicly accessible URL to an image resource such as a jpg or png file.
+      description: |
+        The ImageAsset adds an image to a Clip. The image can be sourced from a URL
+        (`src`) or generated from a text prompt (`prompt`). Exactly one of `src` or
+        `prompt` must be provided.
+
+        - **Source URL:** set `src` to the publicly accessible URL of a jpg or png file.
+        - **Generated:** set `prompt` to describe the image; the engine generates it
+          using the provider chosen by `model` and fills `src` in automatically.
       type: object
       properties:
         type:
@@ -12,14 +17,27 @@
         src:
           description: >-
             The image source URL. The URL must be publicly accessible or include
-            credentials.
+            credentials. Provide either `src` or `prompt`, not both.
           type: string
           minLength: 1
           pattern: '\S'
           example: https://s3-ap-northeast-1.amazonaws.com/my-bucket/image.jpg
+        prompt:
+          description: >-
+            A text prompt to generate the image from. When set without `src`, the
+            engine generates an image and fills `src` automatically. Use `model` to
+            choose the generator.
+          type: string
+          maxLength: 4000
+          example: A serene landscape with a crystal-clear mountain lake at sunrise.
+        model:
+          description: >-
+            The generation model to use when `prompt` is set (e.g. `flux-schnell`).
+            Defaults to the platform's preferred generator if omitted.
+          type: string
+          example: flux-schnell
         crop:
           $ref: "./crop.yaml#/Crop"
       additionalProperties: false
       required:
         - type
-        - src

schemas/imagetovideoasset.yaml

@@ -1,5 +1,11 @@
 ImageToVideoAsset:
-  description: >-
+  deprecated: true
+  description: |
+    **Notice: ImageToVideoAsset is deprecated. Use [VideoAsset](#tocs_videoasset)
+    with `prompt` and `seed` instead.** This type continues to function and is
+    internally rewritten to VideoAsset; no behaviour change for existing
+    integrations.
+
     The ImageToVideoAsset lets you create a video from an image and a text prompt.
   properties:
     type:

schemas/texttoimageasset.yaml

@@ -1,5 +1,10 @@
 TextToImageAsset:
-  description: >-
+  deprecated: true
+  description: |
+    **Notice: TextToImageAsset is deprecated. Use [ImageAsset](#tocs_imageasset)
+    with `prompt` instead.** This type continues to function and is internally
+    rewritten to ImageAsset; no behaviour change for existing integrations.
+
     The TextToImageAsset lets you create a dynamic image from a text prompt.
   properties:
     type:

schemas/texttospeechasset.yaml

@@ -1,5 +1,11 @@
 TextToSpeechAsset:
-  description: >-
+  deprecated: true
+  description: |
+    **Notice: TextToSpeechAsset is deprecated. Use [AudioAsset](#tocs_audioasset)
+    with `prompt` (the spoken text) and `voice` instead.** This type continues to
+    function and is internally rewritten to AudioAsset; no behaviour change for
+    existing integrations.
+
     The TextToSpeechAsset lets you generate a voice over from text using a text-to-speech service.
     The generated audio can be trimmed, faded and have its volume and speed adjusted using the
     same properties available on the AudioAsset.

schemas/videoasset.yaml

@@ -1,8 +1,14 @@
     VideoAsset:
-      description: >-
-        The VideoAsset is used to create video sequences from video files. The
-        src must be a publicly accessible URL to a video resource such as an mp4
-        file.
+      description: |
+        The VideoAsset adds a video to a Clip. The video can be sourced from a URL
+        (`src`) or generated from a text prompt (`prompt`), optionally seeded from a
+        starting image (`seed`). Exactly one of `src` or `prompt` must be provided.
+
+        - **Source URL:** set `src` to the URL of an mp4 (or compatible) video file.
+        - **Generated:** set `prompt` to describe the motion. Optionally set `seed` to
+          a starting image URL (image-to-video). Use `model` to choose the generator
+          (e.g. `luma-ray-3`, `runpod-itv-mini`). The generated `src` is filled in
+          automatically.
       type: object
       properties:
         type:
@@ -13,11 +19,34 @@
         src:
           description: >-
             The video source URL. The URL must be publicly accessible or include
-            credentials.
+            credentials. Provide either `src` or `prompt`, not both.
           type: string
           minLength: 1
           pattern: '\S'
           example: https://s3-ap-northeast-1.amazonaws.com/my-bucket/video.mp4
+        prompt:
+          description: >-
+            A text prompt to generate the video from. When set without `src`, the
+            engine generates a video and fills `src` automatically. Optionally pair
+            with `seed` for image-to-video. Use `model` to choose the generator.
+          type: string
+          maxLength: 4000
+          example: Slowly zoom out and orbit left around the object.
+        seed:
+          description: >-
+            Seed image URL for image-to-video generation. The image is used as the
+            starting frame; `prompt` describes the motion. Has no effect unless
+            `prompt` is set.
+          type: string
+          minLength: 1
+          example: https://s3-ap-northeast-1.amazonaws.com/my-bucket/seed-image.jpg
+        model:
+          description: >-
+            The generation model to use when `prompt` is set (e.g. `luma-ray-3`,
+            `runpod-itv-mini`). Defaults to the platform's preferred generator if
+            omitted.
+          type: string
+          example: luma-ray-3
         transcode:
           type: boolean
           description: >-
@@ -69,7 +98,7 @@
         speed:
           description: >-
             Adjust the playback speed of the video clip between 0 (paused) and 10 (10x normal speed) where 1 is normal
-            speed (defaults to 1). Adjusting the speed will also adjust the duration of the clip and may require you to 
+            speed (defaults to 1). Adjusting the speed will also adjust the duration of the clip and may require you to
             adjust the Clip length. For example, if you set speed to 0.5, the clip will need to be 2x as long to play
             the entire video (i.e. original length / 0.5). If you set speed to 2, the clip will need to be half as long
             to play the entire video (i.e. original length / 2).
@@ -85,4 +114,3 @@
       additionalProperties: false
       required:
         - type
-        - src

scripts/fix-discriminator.cjs

@@ -361,8 +361,45 @@ function addLegacyTextWrapMigrationError(code, zPrefix) {
   return code;
 }
 
+function addSrcOrPromptRefine(code, zPrefix) {
+  const schemaNames = [
+    'imageassetImageAssetSchema',
+    'videoassetVideoAssetSchema',
+    'audioassetAudioAssetSchema',
+  ];
+  const strictMarker = '.strict()';
+  const superRefine = strictMarker + '.superRefine((data, ctx) => {\n' +
+    '  var hasSrc = typeof data.src === "string" && data.src.trim().length > 0;\n' +
+    '  var hasPrompt = typeof data.prompt === "string" && data.prompt.trim().length > 0;\n' +
+    '  if (!hasSrc && !hasPrompt) {\n' +
+    '    ctx.addIssue({\n' +
+    '      code: ' + zPrefix + '.ZodIssueCode.custom,\n' +
+    '      message: "A media asset requires either \\"src\\" (a source URL) or \\"prompt\\" (to generate the asset). Provide one.",\n' +
+    '      path: ["src"],\n' +
+    '    });\n' +
+    '  }\n' +
+    '})';
+
+  for (const schemaName of schemaNames) {
+    const idx = code.indexOf(schemaName + ' = ');
+    if (idx === -1) {
+      console.log("⚠ src/prompt refine: could not find " + schemaName + " (" + zPrefix + ")");
+      continue;
+    }
+    const strictIdx = code.indexOf(strictMarker, idx);
+    if (strictIdx === -1 || strictIdx - idx > 6000) {
+      console.log("⚠ src/prompt refine: no .strict() for " + schemaName + " (" + zPrefix + ")");
+      continue;
+    }
+    code = code.substring(0, strictIdx) + superRefine + code.substring(strictIdx + strictMarker.length);
+    console.log("✓ Added src/prompt refine to " + schemaName + " (" + zPrefix + ")");
+  }
+  return code;
+}
+
 content = addStrictToObjects(content, "z");
 content = addLegacyTextWrapMigrationError(content, "z");
+content = addSrcOrPromptRefine(content, "z");
 
 fs.writeFileSync(zodGenPath, content);
 
@@ -584,6 +621,7 @@ const clipClipSchemaWithFitFilter = exports.clipClipSchema.transform((clip) => {
 
   cjsContent = addStrictToObjects(cjsContent, "zod_1.z");
   cjsContent = addLegacyTextWrapMigrationError(cjsContent, "zod_1.z");
+  cjsContent = addSrcOrPromptRefine(cjsContent, "zod_1.z");
 
   fs.writeFileSync(zodGenCjsPath, cjsContent);
 }
@@ -717,6 +755,7 @@ const clipClipSchemaWithFitFilter = clipClipSchema.transform((clip) => {
 
   jsContent = addStrictToObjects(jsContent, "z");
   jsContent = addLegacyTextWrapMigrationError(jsContent, "z");
+  jsContent = addSrcOrPromptRefine(jsContent, "z");
 
   fs.writeFileSync(zodGenJsPath, jsContent);
 }