Спасибо, что подметили)

In particular, and in contrast to the original Stable Diffusion architecture, we use a heterogeneous distribution of transformer blocks within the UNet: For efficiency reasons, we omit the transformer block at the highest feature level, use 2 and 10 blocks at the lower levels, and remove the lowest level (8× downsampling) in the UNet altogether — see Tab. 1 for a comparison between the architectures of Stable Diffusion 1.x & 2.x and SDXL. We opt for a more powerful pre-trained text encoder that we use for text conditioning. Specifically, we use OpenCLIP ViT-bigG [19] in combination with CLIP ViT-L [34], where we concatenate the penultimate text encoder outputs along the channel-axis [1]. Besides using cross-attention layers to condition the model on the text-input, we follow [30] and additionally condition the model on the pooled text embedding from the OpenCLIP model.

Prior to GPT-4o, you could use Voice Mode to talk to ChatGPT with latencies of 2.8 seconds (GPT-3.5) and 5.4 seconds (GPT-4) on average. To achieve this, Voice Mode is a pipeline of three separate models: one simple model transcribes audio to text, GPT-3.5 or GPT-4 takes in text and outputs text, and a third simple model converts that text back to audio. This process means that the main source of intelligence, GPT-4, loses a lot of information—it can’t directly observe tone, multiple speakers, or background noises, and it can’t output laughter, singing, or express emotion.

With GPT-4o, we trained a single new model end-to-end across text, vision, and audio, meaning that all inputs and outputs are processed by the same neural network. Because GPT-4o is our first model combining all of these modalities, we are still just scratching the surface of exploring what the model can do and its limitations.