videopython 0.28.3__tar.gz → 0.29.1__tar.gz

{videopython-0.28.3 → videopython-0.29.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: videopython
-Version: 0.28.3
+Version: 0.29.1
 Summary: Minimal video generation and processing library.
 Project-URL: Homepage, https://videopython.com
 Project-URL: Repository, https://github.com/bartwojtowicz/videopython/
@@ -27,14 +27,15 @@ Requires-Dist: accelerate>=0.29.2; extra == 'ai'
 Requires-Dist: chatterbox-tts>=0.1.7; extra == 'ai'
 Requires-Dist: demucs>=4.0.0; extra == 'ai'
 Requires-Dist: diffusers>=0.30.0; extra == 'ai'
-Requires-Dist: easyocr>=1.7.0; extra == 'ai'
 Requires-Dist: hf-transfer>=0.1.9; extra == 'ai'
+Requires-Dist: imagehash>=4.3; extra == 'ai'
 Requires-Dist: llama-cpp-python>=0.3.0; extra == 'ai'
 Requires-Dist: numba>=0.61.0; extra == 'ai'
 Requires-Dist: ollama>=0.4.5; extra == 'ai'
 Requires-Dist: openai-whisper>=20240930; extra == 'ai'
 Requires-Dist: pyannote-audio>=4.0.0; extra == 'ai'
 Requires-Dist: pyloudnorm>=0.1.1; extra == 'ai'
+Requires-Dist: qwen-vl-utils>=0.0.10; extra == 'ai'
 Requires-Dist: scikit-learn>=1.3.0; extra == 'ai'
 Requires-Dist: scipy>=1.10.0; extra == 'ai'
 Requires-Dist: sentencepiece>=0.1.99; extra == 'ai'
@@ -56,6 +57,8 @@ Minimal, LLM-friendly Python library for programmatic video editing, processing,
 
 Full documentation: [videopython.com](https://videopython.com)
 
+> **Disclaimer:** This project started as a hand-written hobby project, but most of the code is now produced by LLM agents. Humans still drive direction, approve changes, and own design decisions.
+
 ## Installation
 
 ### 1. Install FFmpeg
@@ -193,10 +196,10 @@ API docs: [Core](https://videopython.com/api/index/) | [Video](https://videopyth
 | Area | Highlights |
 |---|---|
 | **Generation** | `TextToVideo`, `ImageToVideo`, `TextToImage`, `TextToSpeech`, `TextToMusic` |
-| **Understanding** | `AudioToText` (transcription), `AudioClassifier`, `SceneVLM` (visual scene description), `ActionRecognizer` |
+| **Understanding** | `AudioToText` (transcription), `AudioClassifier`, `SceneVLM` (structured visual scene description), `FaceTracker` (per-shot face tracks) |
 | **Scene detection** | `SemanticSceneDetector` (neural scene boundaries) |
 | **Video analysis** | `VideoAnalyzer` - full-pipeline analysis combining multiple AI capabilities |
-| **Transforms** | `FaceTracker`, `FaceTrackingCrop`, `SplitScreenComposite` |
+| **Transforms** | `FaceTrackingCrop`, `SplitScreenComposite` |
 | **Dubbing** | `VideoDubber` - voice cloning and revoicing with timing sync |
 | **Object swapping** | `ObjectSwapper` - detect, segment, and inpaint objects in video |
 

{videopython-0.28.3 → videopython-0.29.1}/README.md

@@ -8,6 +8,8 @@ Minimal, LLM-friendly Python library for programmatic video editing, processing,
 
 Full documentation: [videopython.com](https://videopython.com)
 
+> **Disclaimer:** This project started as a hand-written hobby project, but most of the code is now produced by LLM agents. Humans still drive direction, approve changes, and own design decisions.
+
 ## Installation
 
 ### 1. Install FFmpeg
@@ -145,10 +147,10 @@ API docs: [Core](https://videopython.com/api/index/) | [Video](https://videopyth
 | Area | Highlights |
 |---|---|
 | **Generation** | `TextToVideo`, `ImageToVideo`, `TextToImage`, `TextToSpeech`, `TextToMusic` |
-| **Understanding** | `AudioToText` (transcription), `AudioClassifier`, `SceneVLM` (visual scene description), `ActionRecognizer` |
+| **Understanding** | `AudioToText` (transcription), `AudioClassifier`, `SceneVLM` (structured visual scene description), `FaceTracker` (per-shot face tracks) |
 | **Scene detection** | `SemanticSceneDetector` (neural scene boundaries) |
 | **Video analysis** | `VideoAnalyzer` - full-pipeline analysis combining multiple AI capabilities |
-| **Transforms** | `FaceTracker`, `FaceTrackingCrop`, `SplitScreenComposite` |
+| **Transforms** | `FaceTrackingCrop`, `SplitScreenComposite` |
 | **Dubbing** | `VideoDubber` - voice cloning and revoicing with timing sync |
 | **Object swapping** | `ObjectSwapper` - detect, segment, and inpaint objects in video |
 

{videopython-0.28.3 → videopython-0.29.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "videopython"
-version = "0.28.3"
+version = "0.29.1"
 description = "Minimal video generation and processing library."
 authors = [
     { name = "Bartosz Wójtowicz", email = "bartoszwojtowicz@outlook.com" },
@@ -70,7 +70,6 @@ ai = [
     "scikit-learn>=1.3.0",
     # Detection backends
     "ultralytics>=8.0.0",
-    "easyocr>=1.7.0",
     # Audio classification (AST via transformers - no separate dep needed)
     # Scene detection
     "transnetv2-pytorch>=1.0.5",
@@ -84,6 +83,11 @@ ai = [
     "llama-cpp-python>=0.3.0",
     # Loudness measurement (BS.1770) for dub-vs-source loudness matching (M3)
     "pyloudnorm>=0.1.1",
+    # Vision-language preprocessing for Qwen3.5 (M5) - documented prerequisite
+    # for AutoModelForImageTextToText with image/video chat templates.
+    "qwen-vl-utils>=0.0.10",
+    # Perceptual hashing for SceneVLM frame dedup (M5)
+    "imagehash>=4.3",
 ]
 
 # Required for pip install videopython[ai] - pip uses optional-dependencies, not dependency-groups
@@ -105,7 +109,6 @@ ai = [
     "scikit-learn>=1.3.0",
     # Detection backends
     "ultralytics>=8.0.0",
-    "easyocr>=1.7.0",
     # Audio classification (AST via transformers - no separate dep needed)
     # Scene detection
     "transnetv2-pytorch>=1.0.5",
@@ -119,6 +122,11 @@ ai = [
     "llama-cpp-python>=0.3.0",
     # Loudness measurement (BS.1770) for dub-vs-source loudness matching (M3)
     "pyloudnorm>=0.1.1",
+    # Vision-language preprocessing for Qwen3.5 (M5) - documented prerequisite
+    # for AutoModelForImageTextToText with image/video chat templates.
+    "qwen-vl-utils>=0.0.10",
+    # Perceptual hashing for SceneVLM frame dedup (M5)
+    "imagehash>=4.3",
 ]
 
 [project.urls]
@@ -135,7 +143,6 @@ module = [
     "diffusers", "diffusers.*",
     "ollama", "ollama.*",
     "ultralytics", "ultralytics.*",
-    "easyocr", "easyocr.*",
     "transformers", "transformers.*",
     "transnetv2_pytorch", "transnetv2_pytorch.*",
     "chatterbox", "chatterbox.*",
@@ -146,6 +153,8 @@ module = [
     "cv2", "cv2.*",
     "llama_cpp", "llama_cpp.*",
     "pyloudnorm", "pyloudnorm.*",
+    "qwen_vl_utils", "qwen_vl_utils.*",
+    "imagehash", "imagehash.*",
 ]
 ignore_missing_imports = true
 

{videopython-0.28.3 → videopython-0.29.1}/src/videopython/ai/__init__.py

@@ -2,11 +2,11 @@ from videopython.ai import registry as _ai_registry  # noqa: F401
 
 from .generation import ImageToVideo, TextToImage, TextToMusic, TextToSpeech, TextToVideo
 from .swapping import ObjectSwapper
-from .transforms import FaceTracker, FaceTrackingCrop, SplitScreenComposite
+from .transforms import FaceTrackingCrop, SplitScreenComposite
 from .understanding import (
-    ActionRecognizer,
     AudioClassifier,
     AudioToText,
+    FaceTracker,
     SceneVLM,
     SemanticSceneDetector,
 )
@@ -22,12 +22,10 @@ __all__ = [
     # Understanding
     "AudioToText",
     "AudioClassifier",
+    "FaceTracker",
     "SceneVLM",
-    # Temporal
-    "ActionRecognizer",
     "SemanticSceneDetector",
     # Transforms (AI-powered)
-    "FaceTracker",
     "FaceTrackingCrop",
     "SplitScreenComposite",
     # Swapping

{videopython-0.28.3 → videopython-0.29.1}/src/videopython/ai/dubbing/cache.py

@@ -27,6 +27,8 @@ from dataclasses import dataclass
 from pathlib import Path
 from typing import TYPE_CHECKING, Any
 
+from videopython.ai.understanding.audio import _normalize_vocabulary
+
 if TYPE_CHECKING:
     from videopython.base.audio import Audio
     from videopython.base.text.transcription import Transcription
@@ -37,7 +39,12 @@ logger = logging.getLogger(__name__)
 # Cache schema version. Bump on incompatible changes to any artifact's
 # on-disk format (e.g. TranscriptionSegment field changes that break
 # from_dict). Mismatched cache entries are treated as a miss.
-SCHEMA_VERSION = 1
+#
+# v2 (0.29.1): vocabulary added to transcription_kwargs_hash for M1
+# vocabulary biasing. Pre-v2 transcription artifacts miss on first hit
+# and re-run; translation/TTS artifacts are unaffected (hashed
+# independently and survive).
+SCHEMA_VERSION = 2
 
 # Reserved for M4.3 per-speaker voice library. M3.2 does not write here;
 # documented so future code knows the path is taken.
@@ -126,13 +133,22 @@ class DubCache:
         condition_on_previous_text: bool,
         no_speech_threshold: float,
         logprob_threshold: float | None,
+        vocabulary: list[str] | None = None,
     ) -> str:
+        """Hash captures the kwargs that affect Whisper's output.
+
+        ``vocabulary`` is normalized (case-insensitive dedup, casing
+        preserved) before hashing so trivial reordering/casing
+        differences don't thrash the cache. Defaults to ``None`` so
+        pre-M1 callers keep hashing the same value as before.
+        """
         return _stable_hash(
             whisper_model,
             enable_diarization,
             condition_on_previous_text,
             no_speech_threshold,
             logprob_threshold,
+            *_normalize_vocabulary(vocabulary),
         )
 
     @staticmethod

{videopython-0.28.3 → videopython-0.29.1}/src/videopython/ai/dubbing/dubber.py

@@ -37,6 +37,11 @@ class VideoDubber:
             gate; raise to drop more low-confidence windows.
         logprob_threshold: Forwarded to ``AudioToText``. Whisper's average
             log-probability gate.
+        vocabulary: Forwarded to ``AudioToText``. Optional list of brand
+            names, product names, or proper nouns to bias Whisper's first-
+            window decoder via ``initial_prompt``. Recovers near-mishears
+            (e.g. Klarna → "carna") on brand-monitoring inputs without new
+            model deps.
         strict_quality: When True, the pipeline raises
             :class:`GarbageTranscriptError` before Demucs/translation/TTS run
             if the transcript-quality heuristic returns ``"reject"``. When
@@ -67,6 +72,7 @@ class VideoDubber:
         condition_on_previous_text: bool = False,
         no_speech_threshold: float = 0.6,
         logprob_threshold: float | None = -1.0,
+        vocabulary: list[str] | None = None,
         strict_quality: bool = False,
         translator: TranslatorChoice = "auto",
         cache_dir: str | Path | None = None,
@@ -77,6 +83,7 @@ class VideoDubber:
         self.condition_on_previous_text = condition_on_previous_text
         self.no_speech_threshold = no_speech_threshold
         self.logprob_threshold = logprob_threshold
+        self.vocabulary = vocabulary
         self.strict_quality = strict_quality
         self.translator = translator
         self.cache_dir = cache_dir
@@ -101,6 +108,7 @@ class VideoDubber:
             condition_on_previous_text=self.condition_on_previous_text,
             no_speech_threshold=self.no_speech_threshold,
             logprob_threshold=self.logprob_threshold,
+            vocabulary=self.vocabulary,
             strict_quality=self.strict_quality,
             translator=self.translator,
             cache_dir=self.cache_dir,

{videopython-0.28.3 → videopython-0.29.1}/src/videopython/ai/dubbing/pipeline.py

@@ -170,6 +170,7 @@ class LocalDubbingPipeline:
         condition_on_previous_text: bool = False,
         no_speech_threshold: float = 0.6,
         logprob_threshold: float | None = -1.0,
+        vocabulary: list[str] | None = None,
         strict_quality: bool = False,
         translator: TranslatorChoice = "auto",
         cache_dir: str | Path | None = None,
@@ -180,6 +181,7 @@ class LocalDubbingPipeline:
         self.condition_on_previous_text = condition_on_previous_text
         self.no_speech_threshold = no_speech_threshold
         self.logprob_threshold = logprob_threshold
+        self.vocabulary = vocabulary
         self.strict_quality = strict_quality
         self.translator = translator
         self.cache_dir = Path(cache_dir) if cache_dir is not None else None
@@ -256,6 +258,7 @@ class LocalDubbingPipeline:
                     "condition_on_previous_text": self.condition_on_previous_text,
                     "no_speech_threshold": self.no_speech_threshold,
                     "logprob_threshold": self.logprob_threshold,
+                    "vocabulary": self.vocabulary,
                 },
             )
         return transcription
@@ -406,6 +409,7 @@ class LocalDubbingPipeline:
             condition_on_previous_text=self.condition_on_previous_text,
             no_speech_threshold=self.no_speech_threshold,
             logprob_threshold=self.logprob_threshold,
+            vocabulary=self.vocabulary,
         )
         return src_hash, kwargs_hash
 
@@ -420,6 +424,7 @@ class LocalDubbingPipeline:
             condition_on_previous_text=self.condition_on_previous_text,
             no_speech_threshold=self.no_speech_threshold,
             logprob_threshold=self.logprob_threshold,
+            vocabulary=self.vocabulary,
         )
 
     def _init_translator(self, source_lang: str, target_lang: str) -> None: