| --- |
| license: apache-2.0 |
| datasets: |
| - projectlosangeles/Godzilla-MIDI-Dataset |
| - asigalov61/Mono-Segments |
| - projectlosangeles/Discover-MIDI-Dataset |
| - asigalov61/MIDI-Loops |
| language: |
| - en |
| tags: |
| - Orpheus |
| - MIDI |
| - midi |
| - midi-model |
| - midi-ai |
| - music-ai |
| - music-transformer |
| - SOTA |
| - multi-instrumental |
| - music |
| - loops |
| - bridge |
| - infilling |
| - inpainting |
| - awesome |
| metrics: |
| - accuracy |
| --- |
| |
| # Orpheus Music Transformer |
| ## SOTA 8k multi-instrumental music transformer trained on 2.31M+ high-quality MIDIs |
|
|
|  |
|
|
| *** |
| |
| ## Abstract |
| |
| ### Project Los Angeles is very proud to present **Orpheus Music Transformer**, an efficient, SOTA transformer model for long-form, multi-instrumental music generation. At its core lies highly optimized 748M-parameter autoregressive transformer equipped with Rotary Positional Embeddings (RoPE) and Flash Attention, enabling sequence lengths up to 8 k tokens—sufficient to capture extended musical structures. Trained for three epochs on 2.31 million high-quality MIDI tracks from the Godzilla dataset, our model employs a compact 3-token-per-note and 7-token-per-tri-chord encoding, plus a novel duration-and-velocity-last ordering to enhance expressivity. It leverage PyTorch’s bfloat16 precision and memory-efficient sparse-dense products for accelerated inference on CUDA, and provide a top-*p* sampling filter with adjustable temperature. |
| |
| ### The Gradio interface empowers users to upload seed MIDI files or generate from scratch, tune prime/generation token counts, control randomness (temperature, top-*p*), and optionally append drums or natural “outro” tokens. Generated outputs appear in ten parallel batches with synchronized audio previews and piano-roll plots. Users can iteratively add or remove entire batches to sculpt a final composition, which is rendered back into MIDI and audio via an integrated SoundFont pipeline. Our release demonstrates a seamless blend of state-of-the-art model performance, efficient MIDI tokenization, and user-centric design, fostering rapid exploration of algorithmic composition. |
| |
| *** |
|
|
| ## Models |
|
|
| #### Presented are twelve models: |
|
|
| ### Base models |
|
|
| #### **[Orpheus Music Transformer Large (748) Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Music_Transformer_Large_Trained_Model_43860_steps_0.6682_loss_0.8054_acc.pth)** |
| ##### This is a base model that is capable of music generation/continuation and notes/drums inpainting |
|
|
| #### **[Orpheus Music Transformer Medium (479M) Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Music_Transformer_Trained_Model_96332_steps_0.82_loss_0.748_acc.pth)** |
| ##### This is a base model that is capable of music generation/continuation and notes/drums inpainting |
|
|
| #### **[Orpheus Bridge Music Transformer Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Bridge_Music_Transformer_Trained_Model_19571_steps_0.9396_loss_0.7365_acc.pth)** |
| ##### This is an auxiliary model that is capable of seamless bridge inpainting/infilling in any music composition |
|
|
| ### Fine-tuned base models |
|
|
| #### **[Orpheus Karaoke Music Transformer Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Music_Transformer_Karaoke_Fine_Tuned_Model_2068_steps_0.9833_loss_0.7328_acc.pth)** |
| ##### This is a fine-tuned version of the Orpheus base model that is capable of generating Karaoke lyrics template for any MIDI |
|
|
| #### **[Orpheus LRNO Segments Music Transformer Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Music_Transformer_LRNO_Segments_Fine_Tuned_Model_1986_steps_0.5946_loss_0.8384_acc.pth)** |
| ##### This is a fine-tuned version of the Orpheus base model that is capable of music segmentation and music segments generation |
|
|
| #### **[Orpheus Loops Music Transformer Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Loops_Music_Transformer_Fine_Tuned_Model_3441_steps_0.7715_loss_0.7992_acc.pth)** |
| ##### This is a fine-tuned version of the Orpheus base model that is capable of generating multi-instrumental loops |
|
|
| #### **[Orpheus Children Songs Music Transformer Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Music_Transformer_Children_Songs_Fine_Tuned_Model_60_steps_0.5431_loss_0.838_acc.pth)** |
| ##### This is a fine-tuned version of the Orpheus base model that is great for working with children songs |
|
|
| #### **[Orpheus Mono Melodies Transformer Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Music_Transformer_Mono_Melodies_Fine_Tuned_Model_2844_steps_0.3231_loss_0.9174_acc.pth)** |
| ##### This is a fine-tuned version of the Orpheus base model that is great for working with compositions that have a monophonic melody |
|
|
| ### Aux models |
|
|
| #### **[Orpheus Music Transformer Classifier Model](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/Orpheus_Music_Transformer_Classifier_Trained_Model_23670_steps_0.1837_loss_0.9207_acc.pth)** |
| ##### This is a music classification model that is was trained to distinguish between human music and music generated by Orpheus base model |
|
|
| *** |
| |
| ## Live Hugging Face spaces demos |
| |
| ### **[Orpheus Music Transformer](https://huggingface.co/collections/asigalov61/orpheus-music-transformer-685c3c8e59ed1414c02bb8cd)** |
|
|
| #### If you enjoyed any of the Orpheus Music Transformer demos, please star and duplicate. It helps a lot! 🤗 |
|
|
| *** |
| |
| ## Inference notebooks |
| |
| ### [NEW & SOTA] **[Orpheus Auto-Continuations Generator](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/inference_code/Orpheus_Auto_Continuations_Generator.ipynb)** |
|
|
| ### **[Orpheus Drums Transformer](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/inference_code/Orpheus_Drums_Transformer.ipynb)** |
|
|
| *** |
| |
| ## Training dataset code |
| |
| ### Models were trained on select HQ MIDIs from [Godzilla MIDI Dataset](https://huggingface.co/datasets/projectlosangeles/Godzilla-MIDI-Dataset) |
| ### Please check out [Orpheus Taining Dataset Maker](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/training_data/README.md) notebook for details |
| |
| *** |
|
|
| ## Models training code |
|
|
| ### Please check out [Orpheus Music Transformer Maker](https://huggingface.co/asigalov61/Orpheus-Music-Transformer/blob/main/training_code/README.md) code/notebook for details |
|
|
| *** |
| |
| ### Project Los Angeles |
| ### Tegridy Code 2026 |
