# Modeling Music as a Time-Frequency Image: A 2D Tokenizer for Music Generation

arXiv: https://arxiv.org/abs/2605.15831 · HTML: https://mlsanity.com/paper/2605.15831/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 33 references that are in the map.
- First posted on arXiv 15 May 2026.

## Builds on (most cited first, 25 of 25)

- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Image-to-Image Translation with Conditional Adversarial Networks](https://mlsanity.com/paper/1611.07004/) (2016)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [Perceptual Losses for Real-Time Style Transfer and Super-Resolution](https://mlsanity.com/paper/1603.08155/) (2016)
- [Cosmos World Foundation Model Platform for Physical AI](https://mlsanity.com/paper/2501.03575/) (2025)
- [SoundStream: An End-to-End Neural Audio Codec](https://mlsanity.com/paper/2107.03312/) (2021)
- [High Fidelity Neural Audio Compression](https://mlsanity.com/paper/2210.13438/) (2022)
- [AudioLM: a Language Modeling Approach to Audio Generation](https://mlsanity.com/paper/2209.03143/) (2022)
- [Simple and Controllable Music Generation](https://mlsanity.com/paper/2306.05284/) (2023)
- [MusicLM: Generating Music From Text](https://mlsanity.com/paper/2301.11325/) (2023)
- [High-Fidelity Audio Compression with Improved RVQGAN](https://mlsanity.com/paper/2306.06546/) (2023)
- [BigVGAN: A Universal Neural Vocoder with Large-Scale Training](https://mlsanity.com/paper/2206.04658/) (2022)
- [FMA: A Dataset For Music Analysis](https://mlsanity.com/paper/1612.01840/) (2016)
- [Stable Audio Open](https://mlsanity.com/paper/2407.14358/) (2024)
- [Fast Timing-Conditioned Latent Audio Diffusion](https://mlsanity.com/paper/2402.04825/) (2024)
- [Meta Audiobox Aesthetics: Unified Automatic Quality Assessment for Speech, Music, and Sound](https://mlsanity.com/paper/2502.05139/) (2025)
- [UniAudio: An Audio Foundation Model Toward Universal Audio Generation](https://mlsanity.com/paper/2310.00704/) (2023)
- [The Song Describer Dataset: a Corpus of Audio Captions for Music-and-Language Evaluation](https://mlsanity.com/paper/2311.10057/) (2023)
- [Mel-Band RoFormer for Music Source Separation](https://mlsanity.com/paper/2310.01809/) (2023)
- [APCodec: A Neural Audio Codec with Parallel Amplitude and Phase Spectrum Encoding and Decoding](https://mlsanity.com/paper/2402.10533/) (2024)
- [Spectral Codecs: Improving Non-Autoregressive Speech Synthesis with Spectrogram-Based Audio Codecs](https://mlsanity.com/paper/2406.05298/) (2024)
- [MOSS-Audio-Tokenizer: Scaling Audio Tokenizers for Future Audio Foundation Models](https://mlsanity.com/paper/2602.10934/) (2026)
- [An Independence-promoting Loss for Music Generation with Language Models](https://mlsanity.com/paper/2406.02315/) (2024)
- [STFTCodec: High-Fidelity Audio Compression through Time-Frequency Domain Representation](https://mlsanity.com/paper/2503.16989/) (2025)
- [UniSRCodec: Unified and Low-Bitrate Single Codebook Codec with Sub-Band Reconstruction](https://mlsanity.com/paper/2601.02776/) (2026)

## Built on by (most cited first, 1 of 1)

- [Agogic: Performance-Timed Music Tokens for LLM-Native Text-to-Symbolic-Music Generation](https://mlsanity.com/paper/2608.03999/) (2026)

## Abstract (arXiv)

Autoregressive music generation depends strongly on the audio tokenizer. Existing high-fidelity codecs often use residual multi-codebook quantization, which preserves reconstruction quality but complicates language modeling after sequence flattening, as the residual hierarchy imposes strong sequential dependencies and can amplify error accumulation. We propose BandTok, a generation-oriented 2D Mel-spectrogram tokenizer that represents each frame with Mel-frequency band tokens from a single shared codebook. This design yields a physically interpretable time-frequency token grid with a more independent token structure, making it better suited for autoregressive modeling. BandTok improves reconstruction with a multi-scale PatchGAN objective and EMA codebook updates. We further introduce an autoregressive language model with 2D Rotary Position Embedding (2D RoPE) to preserve temporal and frequency-band structure during generation. Experiments show that BandTok improves over residual-codebook tokenizers and achieves strong results in a data-limited setting. The source code and generation demos for this work are publicly available.