Speech language models (SpeechLMs) accept speech input and produce speech output, allowing for more natural human-computer interaction compared to text-based large language models (LLMs). Traditional approaches for developing SpeechLMs are constrained by the limited availability of unsupervised speech data and parallel speech-text data, which are significantly less abundant compared to text pre-training data, thereby limiting their scalability as LLMs. We propose a novel approach to scaling speech-text pre-training by leveraging large-scale synthetic interleaved data derived from text corpora, eliminating the need for parallel speech-text datasets. Our method efficiently constructs speech-text interleaved data by sampling text spans from existing text corpora and synthesizing corresponding speech spans using a text-to-token model, bypassing the need to generate actual speech. We also employ a supervised speech tokenizer derived from an automatic speech recognition (ASR) model by incorporating a vector-quantized bottleneck into the encoder. This supervised training approach results in discrete speech tokens with strong semantic preservation even at lower sampling rates (e.g. 12.5Hz), while still maintaining speech reconstruction quality. Starting from a pre-trained language model and scaling our pre-training to 1 trillion tokens (with 600B synthetic interleaved speech-text data), we achieve state-of-the-art performance in both speech language modeling and spoken question answering, improving performance on spoken questions tasks from the previous SOTA of 13\% (Moshi) to 31\%. We further demonstrate that by fine-tuning the pre-trained model with speech dialogue data, we can develop an end-to-end spoken chatbot that achieves competitive performance comparable to existing baselines in both conversational abilities and speech quality, even operating exclusively in the speech domain.

This paper proposed a new method for cephalogram landmark location. It firstly employed diffusions in different scales for layered segmentation, making uses of region-homogeneity and edge-saltation, thus the landmark positions could be clearly reflected on edge-maps, i.e. skeleton, of layered diffusion. Secondly, key landmarks were determined via binarization pixel-cliques which were formed by Euclidean distance maps (EDM) on dilated skeleton. To validate the performance of this method, 30 surgery cases were inspected. The comparison of predicted parameters and application values shows that all 9 angle parameters and 4 of 5 distance parameters were correctly calculated, the predicted profiles were similar to the actual contours except one point due to the intrinsic difficulty in labium prediction. Our method took advantages of layered feature preserving and ease landmark extraction, and the latter was also owing to different EDM-Width of pixel-cliques and skeleton points. It is accurate, lifelike, and superior to many similar methods