Merge pull request #6 from FolhaSP/feat/news-video-script-generator

feat: add news video script generator and fix effects

pyproject.toml

@@ -29,6 +29,7 @@ assemblyai = ["assemblyai>=0.34.0"]
 langchain = ["langchain>=0.3.7"]
 haystack = ["haystack-ai>=2.6.1"]
 elevenlabs = ["elevenlabs>=1.9.0"]
+news = ["instructor>=1.6.4", "litellm>=1.52.9"]
 
 [build-system]
 requires = ["hatchling"]

src/mosaico/assets/subtitle.py

@@ -21,7 +21,7 @@ class SubtitleAsset(BaseTextAsset):
         default_factory=lambda: TextAssetParams(
             position=RegionPosition(x="center", y="bottom"),
             font_color=Color("white"),
-            font_size=48,
+            font_size=45,
             stroke_width=1,
             align="center",
             shadow_blur=10,

src/mosaico/script_generators/script.py

@@ -23,8 +23,8 @@ class Shot(BaseModel):
     subtitle: str
     """The subtitle for the shot."""
 
-    media_references: list[int] = Field(..., min_length=1)
-    """The media references for the shot."""
+    media_id: str
+    """The media reference for the shot."""
 
     effects: list[VideoEffectType] = Field(default_factory=list)
     """The effects applied to the shot."""

src/mosaico/speech_synthesizers/openai.py

@@ -2,7 +2,6 @@
 from collections.abc import Sequence
 from typing import Annotated, Any, ClassVar, Literal
 
-from openai import OpenAI
 from pydantic import BaseModel
 from pydantic.fields import Field, PrivateAttr
 from pydantic.functional_validators import model_validator
@@ -45,6 +44,10 @@ def _set_client(self) -> Self:
         """
         Set the OpenAI client.
         """
+        try:
+            from openai import OpenAI
+        except ImportError:
+            raise ImportError("The 'openai' package is required for using the OpenAISpeechSynthesizer.")
         self._client = OpenAI(api_key=self.api_key, base_url=self.base_url, timeout=self.timeout)
         return self
 

src/mosaico/video/project.py

@@ -127,26 +127,26 @@ def from_script_generator(
 
         # Create assets and asset references from the script.
         for shot in script.shots:
-            shot_assets: list[Asset] = [SubtitleAsset.from_data(shot.subtitle)]
-            shot_scene = Scene(description=shot.description).add_asset_references(
-                references=AssetReference.from_asset(shot_assets[0])
-                .with_start_time(shot.start_time)
-                .with_end_time(shot.end_time)
-            )
+            referenced_media = next(m for m in media if m.id == shot.media_id)
+            shot_subtitle = SubtitleAsset.from_data(shot.subtitle)
             shot_effects = [create_effect(effect) for effect in shot.effects]
-
-            for media_ref in shot.media_references:
-                shot_asset = convert_media_to_asset(media[media_ref])
-                shot_asset_ref = (
+            shot_asset = convert_media_to_asset(referenced_media)
+            shot_scene = (
+                Scene(description=shot.description)
+                .add_asset_references(
+                    AssetReference.from_asset(shot_subtitle)
+                    .with_start_time(shot.start_time)
+                    .with_end_time(shot.end_time)
+                )
+                .add_asset_references(
                     AssetReference.from_asset(shot_asset)
                     .with_start_time(shot.start_time)
                     .with_end_time(shot.end_time)
-                    .with_effects(shot_effects)
+                    .with_effects(shot_effects if shot_asset.type == "image" else [])
                 )
-                shot_scene = shot_scene.add_asset_references(shot_asset_ref)
-                shot_assets.append(shot_asset)
+            )
 
-            project = project.add_assets(shot_assets).add_timeline_events(shot_scene)
+            project = project.add_assets(shot_asset).add_assets(shot_subtitle).add_timeline_events(shot_scene)
 
         return project
 

src/mosaico/video/rendering.py

@@ -16,7 +16,6 @@
 
 if TYPE_CHECKING:
     from mosaico.assets.types import Asset
-    from mosaico.effects.protocol import Effect
     from mosaico.types import FrameSize
     from mosaico.video.project import VideoProject
     from mosaico.video.types import TimelineEvent
@@ -35,21 +34,19 @@ def render_video(project: VideoProject, output_dir: str | Path, *, overwrite: bo
     output_path = output_dir / f"{project.config.title}.mp4"
 
     if not output_dir.exists():
-        raise FileNotFoundError(f"Output directory does not exist: {output_dir}")
+        msg = f"Output directory does not exist: {output_dir}"
+        raise FileNotFoundError(msg)
 
     if output_path.exists() and not overwrite:
-        raise FileExistsError(f"Output file already exists: {output_path}")
+        msg = f"Output file already exists: {output_path}"
+        raise FileExistsError(msg)
 
     video_clips = []
     audio_clips = []
 
     for event in project.timeline:
         event_asset_ref_pairs = _get_event_assets_and_refs(event, project)
-        event_video_clips, event_audio_clips = _render_event_clips(
-            asset_and_ref_pairs=event_asset_ref_pairs,
-            video_resolution=project.config.resolution,
-            effects=getattr(event, "effects", []),
-        )
+        event_video_clips, event_audio_clips = _render_event_clips(event_asset_ref_pairs, project.config.resolution)
         video_clips.extend(event_video_clips or [])
         audio_clips.extend(event_audio_clips or [])
 
@@ -99,7 +96,7 @@ def _get_event_asset_refs(event: TimelineEvent) -> list[AssetReference]:
 
 
 def _render_event_clips(
-    asset_and_ref_pairs: Sequence[tuple[Asset, AssetReference]], video_resolution: FrameSize, effects: Sequence[Effect]
+    asset_and_ref_pairs: Sequence[tuple[Asset, AssetReference]], video_resolution: FrameSize
 ) -> tuple[list[VideoClip], list[AudioClip]]:
     """
     Compose a video clip from the given assets.
@@ -108,7 +105,7 @@ def _render_event_clips(
     video_clips = []
 
     for asset, asset_ref in asset_and_ref_pairs:
-        clip = make_clip(asset, asset_ref.duration, video_resolution, effects)
+        clip = make_clip(asset, asset_ref.duration, video_resolution, asset_ref.effects)
         clip = clip.set_start(asset_ref.start_time)
 
         if hasattr(asset.params, "z_index"):

tests/integrations/langchain/test_script_generator.py

@@ -15,7 +15,11 @@
 
 @pytest.fixture
 def sample_media():
-    return [Media.from_data("Test video content", metadata={"description": "A test video", "credit": "Test Author"})]
+    return [
+        Media.from_data(
+            "Test video content", id="media_1", metadata={"description": "A test video", "credit": "Test Author"}
+        )
+    ]
 
 
 @pytest.fixture
@@ -30,7 +34,7 @@ def sample_script():
                 start_time=0.0,
                 end_time=5.0,
                 subtitle="Test subtitle",
-                media_references=[0],
+                media_id="media_1",
             )
         ],
     )
@@ -76,7 +80,7 @@ def generate(self, media: Sequence[Media], **kwargs) -> ShootingScript:
                     start_time=0.0,
                     end_time=5.0,
                     subtitle="Test",
-                    media_references=[0],
+                    media_id=media[0].id,
                 )
             ],
         )

tests/video/test_project.py

@@ -218,15 +218,15 @@ def generate(self, media: Sequence[Media], **kwargs: Any) -> ShootingScript:
                     start_time=0,
                     end_time=5,
                     subtitle="Hello world",
-                    media_references=[0],
+                    media_id=media[0].id,
                 ),
                 Shot(
                     number=2,
                     description="Shot 2",
                     start_time=5,
                     end_time=10,
                     subtitle="This is a test",
-                    media_references=[1],
+                    media_id=media[1].id,
                 ),
             ],
         )