Dolby Laboratories

A user prompt for a diffusion model to generate an image is received. Noisy latent features generated by denoising layers of the diffusion model are collected based on the user prompt and are to generate a pre-customized version of the image. Pre-customized lighting properties of the pre-customized version of the image are predicted based on the noisy latent features. Light source properties of a light source to be logically applied to generate a customized version of the image are determined. The light source properties of the light source are applied to the pre-customized lighting properties of the pre-customized version of the image to generate customized lighting properties of the customized version of the image. A loss computed from the customized lighting properties and the pre-customized lighting properties is used to guide the diffusion model to generate the customized version of the image.

Methods, apparatus, programs, and storage media for enhancing speech signals based on a neural network are provided. The method includes receiving a speech signal and a corresponding reference speech signal. A first set of band features is extracted from the speech signal and a second set of band features is extracted from the reference speech signal, wherein each feature in the first set of band features corresponds to a frequency band of the speech signal and each feature in the second set of band features corresponds to a frequency band of the reference speech signal. The first set of band features is split into a low frequency band part and high frequency band part. Band features corresponding to the high frequency band part are mapped to a reduced number of high frequency band features. The low frequency band part and the reduced number of high frequency band features are combined into a third set of band features. A first mask is determined by inputting the third set of band features into a speech enhancement neural subnetwork. The first mask is split into a low frequency mask part and high frequency mask part. Entries corresponding to the high frequency mask part are mapped to an increased number of entries. The low frequency mask part and the increased number of entries are combined into a second mask. A fourth set of band features is determined based on the second mask and the speech signal. A loss function is evaluated based on the second set of band features and the fourth set of band features. Weights of the speech enhancement neural subnetwork are updated based on a value of the evaluated loss function.

This disclosure falls into the field of audio coding, in particular it is related to the field of providing a framework for providing loudness consistency among differing audio output signals. In particular, the disclosure relates to methods, computer program products and apparatus for encoding and decoding of audio data bitstreams in order to attain a desired loudness level of an output audio signal.

The positions of a plurality of speakers at a media consumption site are determined. Audio information in an object-based format is received. Gain adjustment value for a sound content portion in the object-based format may be determined based on the position of the sound content portion and the positions of the plurality of speakers. Audio information in a ring-based channel format is received. Gain adjustment value for each ring-based channel in a set of ring-based channels may be determined based on the ring to which the ring-based channel belongs and the positions of the speakers at a media consumption site.

An audio processing unit, comprising a buffer memory (201) configured to store at least one frame of an encoded audio bitstream which includes audio data and a metadata container having a header and one or more metadata payloads after the header, a parser (205) coupled to the buffer memory, and an audio decoder (202) coupled to the parser. The one or more metadata payloads include dynamic range compression (DRC) metadata including DRC control values indicative of a compression profile assumed by an encoder to generate said DRC control values. The audio decoder is configured to perform dynamic range compression on decoded audio data generated by decoding said at least some of the audio data, using the DRC metadata.

The present document describes a method (500) for generating a bitstream (101), wherein the bitstream (101) comprises a sequence of superframes (400) for a sequence of frames of an immersive audio signal (111). The method (500) comprises, repeatedly for the sequence of superframes (400), inserting (501) coded audio data (206) for one or more frames of one or more downmix channel signals (203) derived from the immersive audio signal (111), into data fields (411, 421, 412, 422) of a superframe (400); and inserting (502) metadata (202, 205) for reconstructing one or more frames of the immersive audio signal (111) from the coded audio data (206), into a metadata field (403) of the superframe (400).

Many portable playback devices cannot decode and playback encoded audio content having wide bandwidth and wide dynamic range with consistent loudness and intelligibility unless the encoded audio content has been prepared specially for these devices. This problem can be overcome by including with the encoded content some metadata that specifies a suitable dynamic range compression profile by either absolute values or differential values relative to another known compression profile. A playback device may also adaptively apply gain and limiting to the playback audio. Implementations in encoders, in transcoders and in decoders are disclosed.

A quantization parameter signalling mechanism for both SDR and HDR content in video coding is described using two approaches. The first approach is to send the user-defined Qpc table directly in high level syntax. This leads to more flexible and efficient QP control for future codec development and video content coding. The second approach is to signal luma and chroma QPs independently. This approach eliminates the need for Qp_C tables and removes the dependency of chroma quantization parameter on luma QP.

Disclosed is an audio signal encoding/decoding method that uses an encoding downmix strategy applied at an encoder that is different than a decoding re-mix/upmix strategy applied at a decoder. Based on the type of downmix coding scheme, the method comprises: computing input downmixing gains to be applied to the input audio signal to construct a primary downmix channel; determining downmix scaling gains to scale the primary downmix channel; generating prediction gains based on the input audio signal, the input downmixing gains and the downmix scaling gains; determining residual channel(s) from the side channels by using the primary downmix channel and the prediction gains to generate side channel predictions and subtracting the side channel predictions from the side channels; determining decorrelation gains based on energy in the residual channels; encoding the primary downmix channel, the residual channel(s), the prediction gains and the decorrelation gains; and sending the bitstream to a decoder.

Encoding/decoding an immersive voice and audio services (IVAS) bitstream comprises: encoding/decoding a coding mode indicator in a common header (CH) section of an IVAS bitstream, encoding/decoding a mode header or tool header in the tool header (TH) section of the bitstream, the TH section following the CH section, encoding/decoding a metadata payload in a metadata payload (MDP) section of the bitstream, the MDP section following the CH section, encoding/decoding an enhanced voice services (EVS) payload in an EVS payload (EP) section of the bitstream, the EP section following the CH section, and on the encoder side, storing or streaming the encoded bitstream, and on the decoder side, controlling an audio decoder based on the coding mode, the tool header, the EVS payload, and the metadata payload or storing a representation of same.