Spaces:
Running
Running
| title: README | |
| emoji: 馃摎 | |
| colorFrom: blue | |
| colorTo: gray | |
| sdk: static | |
| pinned: false | |
| # MB Mel Dynamics | |
| AI/ML services company in Lithuania. We prepare training data, fine-tune visual and audio models, and deploy them into working systems. | |
| **Website:** [meldynamics.com](https://meldynamics.com) 路 **GitHub:** [MELDYNAMICS](https://github.com/MELDYNAMICS) | |
| ## What we publish here | |
| Open datasets and models built on public source material. Each card states what we contributed and credits the original creators. | |
| | Repository | Type | Our contribution | Source material | | |
| |---|---|---|---| | |
| | [polish-metrical-htr-experimental](https://huggingface.co/meldynamics/polish-metrical-htr-experimental) | Model, experimental | Fine-tuned TrOCR line recognizer for 19th-century Polish metrical registers | Base model: Kansallisarkisto/multicentury-htr-model | | |
| | [liepa-tts](https://huggingface.co/datasets/meldynamics/liepa-tts) | Speech dataset | Recovery, alignment, restructuring, and stress enrichment | Vilnius University, LIEPA project | | |
| | [lt-stressed-corpus](https://huggingface.co/datasets/meldynamics/lt-stressed-corpus) | Text dataset | Word-stress enrichment and derived dataset engineering | MATAS v1.0, ALKSNIS v3.0 | | |
| | [liepa-asr](https://huggingface.co/datasets/meldynamics/liepa-asr) | Speech corpus | Hugging Face edition | Vilnius University, LIEPA project | | |
| | [liepa-2](https://huggingface.co/datasets/meldynamics/liepa-2) | Speech corpus | Hugging Face edition | Vilnius University, LIEPA 2 project | | |
| | [liepa-3](https://huggingface.co/datasets/meldynamics/liepa-3) | Speech corpus | Hugging Face edition | Vytautas Magnus University and Vilnius University | | |
| The LIEPA recordings were created by the universities listed above; we did not collect them. The handwriting model is a research model evaluated against automatic teacher transcriptions, not human ground truth. | |
| ## Services | |
| - **Visual model fine-tuning:** adapting vision and document models to your images and categories. | |
| - **Audio model fine-tuning:** speech recognition and synthesis for specific languages, speakers, or recording conditions. | |
| - **Dataset collection and management:** collection, cleaning, labelling, alignment, provenance, and versioning. | |
| - **Deployment:** packaging models as services or batch jobs in your environment. | |
| To discuss a project, use the contact form at [meldynamics.com](https://meldynamics.com/#contact). | |
| --- | |
| MB Mel Dynamics 路 Company code 306606584 路 Lithuania 路 Established 2023 | |