GARAGEM 2.0: General Automotive Real and Artificial speech corpus for Garage Environments and Maintenance in brazilian portuguese

General Description This dataset is a collection of audio recordings focused on the automotive workshop domain, consisting of a real dataset extracted from YouTube and four synthetic datasets generated by different text-to-speech APIs or open source models. The total dataset contains approximately 61 hours of audio. Dataset Structure The dataset is organized into five main subsets: 1. YouTube Dataset (Real) Contains 4,412 audio segments (9 hours and 45 minutes) extracted from YouTube channels focused on automotive workshop topics. Split:- Training: 3,121 segments (6 hours and 51 minutes)- Test: 1,291 segments (2 hours and 54 minutes) Metadata Files:- train_metadata_youtube.csv- test_metadata_youtube.csv Metadata columns:- url: YouTube video URL- start_timestamp: Segment start time- end_timestamp: Segment end time- human_transcription: Human-annotated transcription- speaker: Speaker annotation 2. Google Cloud Dataset (Synthetic) Contains 6,367 synthetic audio segments generated by Google Cloud API [1]. Format: WAV, 16 KHz Available Metadata Files:- metadata_google_cloud_100.csv: 100% of audios (11 hours and 15 minutes)- metadata_google_cloud_50.csv: 50% of audios (5 hours and 36 minutes)- metadata_google_cloud_20.csv: 20% of audios (2 hours and 14 minutes)- metadata_google_cloud_10.csv: 10% of audios (1 hour and 7 minutes)- metadata_google_cloud_5.csv: 5% of audios (33 minutes) Metadata columns:- audio_filename: Audio file- transcription: Transcription used to generate the audio- speaker: Speaker voice defined in generation configuration 3. Gemini Dataset (Synthetic) Contains 6,367 synthetic audio segments generated by Google API with Gemini assistance [2]. Format: WAV, 16 KHz Available Metadata Files:- metadata_gemini_100.csv: 100% of audios (12 hours and 49 minutes)- metadata_gemini_50.csv: 50% of audios (6 hours and 24 minutes)- metadata_gemini_20.csv: 20% of audios (2 hours and 33 minutes)- metadata_gemini_10.csv: 10% of audios (1 hour and 16 minutes)- metadata_gemini_5.csv: 5% of audios (37 minutes) Metadata columns:- audio_filename: Audio file- transcription: Transcription used to generate the audio- speaker: Speaker voice defined in generation configuration 4. F5 Dataset (Synthetic) Contains 6,367 synthetic audio segments generated by F5 TTS [3]. Format: WAV, 16 KHz Available Metadata Files:- metadata_f5_100.csv: 100% of audios (12 hours and 7 minutes)- metadata_f5_50.csv: 50% of audios (6 hours and 2 minutes)- metadata_f5_20.csv: 20% of audios (2 hours and 24 minutes)- metadata_f5_10.csv: 10% of audios (1 hour and 12 minutes)- metadata_f5_5.csv: 5% of audios (35 minutes) Metadata columns:- audio_filename: Audio file- transcription: Transcription used to generate the audio- base_dataset: source dataset used to obtain original audios, i.e., tedx [4] or cml [5]- base_audio: original audio used to generate synthetic samples 5. XTTS Dataset (Synthetic) Contains 6,367 synthetic audio segments generated by XTTS [6]. Format: WAV, 16 KHz Available Metadata Files:- metadata_xtts_100.csv: 100% of audios (15 hours and 19 minutes)- metadata_xtts_50.csv: 50% of audios (7 hours and 38 minutes)- metadata_xtts_20.csv: 20% of audios (3 hours and 3 minutes)- metadata_xtts_10.csv: 10% of audios (1 hour and 32 minutes)- metadata_xtts_5.csv: 5% of audios (45 minutes) Metadata columns:- audio_filename: Audio file- transcription: Transcription used to generate the audio- base_dataset: source dataset used to obtain original audios, i.e., tedx [4] or cml [5]- base_audio: original audio used to generate synthetic samples Automotive Domain Dictionary The dataset includes a dictionary file (`dictionary.csv`) containing automotive workshop terminology used for generating synthetic sentences. File: dictionary.csv Columns:- term: Automotive workshop term- meaning: Definition of the term (when necessary to assist in generation)- source: Term category Term Distribution:- Automotive parts: 330 terms- Generic term: 142 terms- Car model: 136 terms- Automotive brands: 113 terms Total terms: 721 This dictionary was used as the basis for generating the synthetic sentences present in the dataset, ensuring domain-specific vocabulary coverage. File Organization garagem_v2/├── dictionary.csv├── youtube/│ ├── train_metadata_youtube.csv│ └── test_metadata_youtube.csv├── google_cloud/│ ├── audios_google_cloud/│ ├── metadata_google_cloud_100.csv│ ├── metadata_google_cloud_50.csv│ ├── metadata_google_cloud_20.csv│ ├── metadata_google_cloud_10.csv│ └── metadata_google_cloud_5.csv├── gemini/│ ├── audios_gemini/│ ├── metadata_gemini_100.csv│ ├── metadata_gemini_50.csv│ ├── metadata_gemini_20.csv│ ├── metadata_gemini_10.csv│ └── metadata_gemini_5.csv├── f5/│ ├── audios_f5/│ ├── metadata_f5_100.csv│ ├── metadata_f5_50.csv│ ├── metadata_f5_20.csv│ ├── metadata_f5_10.csv│ └── metadata_f5_5.csv└── xtts/ ├── audios_xtts/ ├── metadata_xtts_100.csv ├── metadata_xtts_50.csv ├── metadata_xtts_20.csv ├── metadata_xtts_10.csv └── metadata_xtts_5.csv Technical Specifications Audio Format (Synthetic Datasets):- Format: WAV- Sample Rate: 16 KHz Metadata File Format:- Format: CSV- Encoding: UTF-8 Recommended Use This dataset is suitable for:- Training automatic speech recognition (ASR) models- Research in speech processing in the automotive workshop domain- Comparing real and proportions of synthetic data- Domain-specific vocabulary analysis Citation If you use this dataset in your research, please cite appropriately. License Creative Commons Attribution 4.0 International References [1]: https://docs.cloud.google.com/text-to-speech/docs/chirp3-hd[2]: https://docs.cloud.google.com/text-to-speech/docs/gemini-tts[3]: https://huggingface.co/Tharyck/multispeaker-ptbr-f5tts[4]: https://www.openslr.org/100[5]: https://www.openslr.org/146/[6]: https://huggingface.co/coqui/XTTS-v2