lyrics-transcriber 0.12.9__tar.gz → 0.14.0__tar.gz

{lyrics_transcriber-0.12.9 → lyrics_transcriber-0.14.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: lyrics-transcriber
-Version: 0.12.9
+Version: 0.14.0
 Summary: Automatically create synchronised lyrics files in ASS and MidiCo LRC formats with word-level timestamps, using Whisper and lyrics from Genius and Spotify
 Home-page: https://github.com/karaokenerds/python-lyrics-transcriber
 License: MIT
@@ -13,21 +13,22 @@ Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
-Requires-Dist: Cython (>=0,<1)
-Requires-Dist: dtw-python (>=1,<2)
-Requires-Dist: llvmlite (>=0,<1)
-Requires-Dist: lyricsgenius (>=3,<4)
-Requires-Dist: numba (>=0.57,<0.58)
-Requires-Dist: numpy (>=1,<2)
-Requires-Dist: onnx (>=1,<2)
-Requires-Dist: onnxruntime (>=1,<2)
+Requires-Dist: Cython (>=0)
+Requires-Dist: dtw-python (>=1)
+Requires-Dist: llvmlite (>=0)
+Requires-Dist: lyricsgenius (>=3)
+Requires-Dist: numba (>=0.57)
+Requires-Dist: numpy (>=1)
+Requires-Dist: onnx (>=1)
+Requires-Dist: onnxruntime (>=1)
 Requires-Dist: openai (>=1,<2)
-Requires-Dist: openai-whisper (==20231117)
-Requires-Dist: python-slugify (>=8,<9)
-Requires-Dist: syrics (>=0,<1)
-Requires-Dist: torch (>1)
-Requires-Dist: tqdm (>=4,<5)
-Requires-Dist: whisper-timestamped (>=1,<2)
+Requires-Dist: openai-whisper (>=20231117)
+Requires-Dist: python-slugify (>=8)
+Requires-Dist: syrics (>=0)
+Requires-Dist: torch (>=1)
+Requires-Dist: tqdm (>=4)
+Requires-Dist: transformers (>=4)
+Requires-Dist: whisper-timestamped (>=1)
 Project-URL: Documentation, https://github.com/karaokenerds/python-lyrics-transcriber/blob/main/README.md
 Project-URL: Repository, https://github.com/karaokenerds/python-lyrics-transcriber
 Description-Content-Type: text/markdown

{lyrics_transcriber-0.12.9 → lyrics_transcriber-0.14.0}/lyrics_transcriber/transcriber.py

@@ -29,7 +29,7 @@ class LyricsTranscriber:
         log_level=logging.DEBUG,
         log_formatter=None,
         transcription_model="medium",
-        llm_model="gpt-4-1106-preview",
+        llm_model="gpt-4o",
         llm_prompt_matching="lyrics_transcriber/llm_prompts/llm_prompt_lyrics_matching_andrew_handwritten_20231118.txt",
         llm_prompt_correction="lyrics_transcriber/llm_prompts/llm_prompt_lyrics_correction_andrew_handwritten_20231118.txt",
         render_video=False,
@@ -66,7 +66,15 @@ class LyricsTranscriber:
         self.llm_model = llm_model
         self.llm_prompt_matching = llm_prompt_matching
         self.llm_prompt_correction = llm_prompt_correction
+
         self.openai_client = OpenAI()
+
+        # Uncomment for local models e.g. with ollama
+        # self.openai_client = OpenAI(
+        #     base_url="http://localhost:11434/v1",
+        #     api_key="ollama",
+        # )
+
         self.openai_client.log = self.log_level
 
         self.render_video = render_video
@@ -190,6 +198,9 @@ class LyricsTranscriber:
             online_lyrics_text_key = f"{online_lyrics_source}_lyrics_text"
             online_lyrics_filepath_key = f"{online_lyrics_source}_lyrics_filepath"
 
+            if online_lyrics_text_key not in self.outputs:
+                continue
+
             data_input_str = (
                 f'Data input 1:\n{self.outputs["transcribed_lyrics_text"]}\nData input 2:\n{self.outputs[online_lyrics_text_key]}\n'
             )
@@ -274,7 +285,7 @@ class LyricsTranscriber:
         total_segments = len(self.outputs["transcription_data_dict"]["segments"])
         self.logger.info(f"Beginning correction using LLM, total segments: {total_segments}")
 
-        with open(self.outputs["llm_transcript_filepath"], "a", buffering=1) as llm_transcript_file:
+        with open(self.outputs["llm_transcript_filepath"], "a", buffering=1, encoding="utf-8") as llm_transcript_file:
             self.logger.debug(f"writing LLM chat instructions: {self.outputs['llm_transcript_filepath']}")
 
             llm_transcript_header = f"--- SYSTEM instructions passed in for all segments ---:\n\n{system_prompt}\n"
@@ -370,7 +381,7 @@ class LyricsTranscriber:
             self.logger.info(f'Successfully processed correction for all {len(corrected_lyrics_dict["segments"])} lyrics segments')
 
             self.logger.debug(f"writing corrected lyrics data JSON filepath: {corrected_lyrics_data_json_cache_filepath}")
-            with open(corrected_lyrics_data_json_cache_filepath, "w") as corrected_lyrics_data_json_cache_file:
+            with open(corrected_lyrics_data_json_cache_filepath, "w", encoding="utf-8") as corrected_lyrics_data_json_cache_file:
                 corrected_lyrics_data_json_cache_file.write(json.dumps(corrected_lyrics_dict, indent=4))
 
         self.outputs["corrected_lyrics_data_filepath"] = corrected_lyrics_data_json_cache_filepath
@@ -388,8 +399,11 @@ class LyricsTranscriber:
             },
         }
 
-        input_cost = price_dollars_per_1000_tokens[self.llm_model]["input"] * (self.outputs["llm_token_usage"]["input"] / 1000)
-        output_cost = price_dollars_per_1000_tokens[self.llm_model]["output"] * (self.outputs["llm_token_usage"]["output"] / 1000)
+        input_price = price_dollars_per_1000_tokens.get(self.llm_model, {"input": 0, "output": 0})["input"]
+        output_price = price_dollars_per_1000_tokens.get(self.llm_model, {"input": 0, "output": 0})["output"]
+
+        input_cost = input_price * (self.outputs["llm_token_usage"]["input"] / 1000)
+        output_cost = output_price * (self.outputs["llm_token_usage"]["output"] / 1000)
 
         self.outputs["llm_costs_usd"]["input"] = round(input_cost, 3)
         self.outputs["llm_costs_usd"]["output"] = round(output_cost, 3)
@@ -405,7 +419,7 @@ class LyricsTranscriber:
             self.outputs["corrected_lyrics_text"] = ""
 
             self.logger.debug(f"writing lyrics plain text to corrected_lyrics_text_filepath: {corrected_lyrics_text_filepath}")
-            with open(corrected_lyrics_text_filepath, "w") as f:
+            with open(corrected_lyrics_text_filepath, "w", encoding="utf-8") as f:
                 for corrected_segment in self.outputs["corrected_lyrics_data_dict"]["segments"]:
                     self.outputs["corrected_lyrics_text"] += corrected_segment["text"].strip() + "\n"
                     f.write(corrected_segment["text".strip()] + "\n")
@@ -452,7 +466,7 @@ class LyricsTranscriber:
             self.logger.debug(
                 f"writing lyrics data JSON to spotify_lyrics_data_json_cache_filepath: {spotify_lyrics_data_json_cache_filepath}"
             )
-            with open(spotify_lyrics_data_json_cache_filepath, "w") as f:
+            with open(spotify_lyrics_data_json_cache_filepath, "w", encoding="utf-8") as f:
                 f.write(spotify_lyrics_json)
         except Exception as e:
             self.logger.warn(f"caught exception while attempting to fetch from spotify: ", e)
@@ -472,7 +486,7 @@ class LyricsTranscriber:
             self.outputs["spotify_lyrics_text"] = ""
 
             self.logger.debug(f"writing lyrics plain text to spotify_lyrics_text_filepath: {spotify_lyrics_text_filepath}")
-            with open(spotify_lyrics_text_filepath, "w") as f:
+            with open(spotify_lyrics_text_filepath, "w", encoding="utf-8") as f:
                 for line in lines:
                     self.outputs["spotify_lyrics_text"] += line["words"].strip() + "\n"
                     f.write(line["words"].strip() + "\n")
@@ -504,7 +518,7 @@ class LyricsTranscriber:
         lyrics = self.clean_genius_lyrics(song.lyrics)
 
         self.logger.debug(f"writing clean lyrics to genius_lyrics_cache_filepath: {genius_lyrics_cache_filepath}")
-        with open(genius_lyrics_cache_filepath, "w") as f:
+        with open(genius_lyrics_cache_filepath, "w", encoding="utf-8") as f:
             f.write(lyrics)
 
         self.outputs["genius_lyrics_filepath"] = genius_lyrics_cache_filepath
@@ -558,7 +572,7 @@ class LyricsTranscriber:
 
         lrc_filename = self.outputs["midico_lrc_filepath"]
         self.logger.debug(f"writing midico formatted word timestamps to LRC file: {lrc_filename}")
-        with open(lrc_filename, "w") as f:
+        with open(lrc_filename, "w", encoding="utf-8") as f:
             f.write("[re:MidiCo]\n")
             for segment in self.outputs["corrected_lyrics_data_dict"]["segments"]:
                 for i, word in enumerate(segment["words"]):
@@ -752,7 +766,7 @@ class LyricsTranscriber:
             self.outputs["transcribed_lyrics_text"] = ""
 
             self.logger.debug(f"writing lyrics plain text to transcribed_lyrics_text_filepath: {transcribed_lyrics_text_filepath}")
-            with open(transcribed_lyrics_text_filepath, "w") as f:
+            with open(transcribed_lyrics_text_filepath, "w", encoding="utf-8") as f:
                 for segment in self.outputs["transcription_data_dict"]["segments"]:
                     self.outputs["transcribed_lyrics_text"] += segment["text"] + "\n"
                     f.write(segment["text"].strip() + "\n")

{lyrics_transcriber-0.12.9 → lyrics_transcriber-0.14.0}/lyrics_transcriber/utils/cli.py

@@ -2,7 +2,6 @@
 import argparse
 import logging
 import pkg_resources
-from lyrics_transcriber import LyricsTranscriber
 
 
 def main():
@@ -66,8 +65,8 @@ def main():
 
     parser.add_argument(
         "--llm_model",
-        default="gpt-4-1106-preview",
-        help="Optional: LLM model to use (currently only supports OpenAI chat completion models, e.g. gpt-4-1106-preview). Default: gpt-3.5-turbo-1106",
+        default="gpt-4o",
+        help="Optional: LLM model to use (currently only supports OpenAI chat completion compatible models",
     )
 
     parser.add_argument(
@@ -110,6 +109,9 @@ def main():
 
     logger.debug("Loading LyricsTranscriber class")
 
+    # Lazy load this class so help output is printed quickly rather than waiting for heavy libraries to load
+    from lyrics_transcriber import LyricsTranscriber
+
     transcriber = LyricsTranscriber(
         args.audio_filepath,
         genius_api_token=args.genius_api_token,

{lyrics_transcriber-0.12.9 → lyrics_transcriber-0.14.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "lyrics-transcriber"
-version = "0.12.9"
+version = "0.14.0"
 description = "Automatically create synchronised lyrics files in ASS and MidiCo LRC formats with word-level timestamps, using Whisper and lyrics from Genius and Spotify"
 authors = ["Andrew Beveridge <andrew@beveridge.uk>"]
 license = "MIT"
@@ -13,21 +13,22 @@ documentation = "https://github.com/karaokenerds/python-lyrics-transcriber/blob/
 
 [tool.poetry.dependencies]
 python = ">=3.9"
-Cython = "^0"
-dtw-python = "^1"
-llvmlite = "^0"
-numba = "^0.57"
-numpy = "^1"
-onnx = "^1"
-onnxruntime = "^1"
-torch = ">1"
-tqdm = "^4"
-lyricsgenius = "^3"
-python-slugify = "^8"
-syrics = "^0"
+Cython = ">=0"
+dtw-python = ">=1"
+llvmlite = ">=0"
+numba = ">=0.57"
+numpy = ">=1"
+onnx = ">=1"
+onnxruntime = ">=1"
+torch = ">=1"
+tqdm = ">=4"
+lyricsgenius = ">=3"
+python-slugify = ">=8"
+syrics = ">=0"
 openai = "^1"
-openai-whisper = "20231117"
-whisper-timestamped = "^1"
+openai-whisper = ">=20231117"
+transformers = ">=4"
+whisper-timestamped = ">=1"
 # Note: after adding openai-whisper and whisper-timestamped with poetry lock, I then removed all traces of triton
 # from poetry.lock before running poetry install, as triton doesn't support macOS but isn't actually needed for whisper.
 # This was the only way I was able to get a working cross-platform build published to PyPI.