# US 2026/0100183 A1 · Method and system for producing synthesized speech digital audio content

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2026/0100183 A1
- Document ID: US20260100183A1
- Status: Published application
- Kind: A1
- Application number: 19/113,174
- Filed: 2023-09-27
- Published: 2026-04-09
- Earliest claimed: 2022-09-27
- Assignees: Voiseed S.R.L.
- Applicants: Voiseed S.R.L.
- Inventors: Lorenzo TARANTINO
- Web page: https://patentgrep.com/patent/US20260100183A1

## CPC

- G10L 13/033: Voice editing, e.g. manipulating the voice of the synthesiser
- G10L 13/047: Architecture of speech synthesisers
- G10L 13/0335: {Pitch control}
- G10L 13/06: Elementary speech units used in speech synthesisers; Concatenation rules
- G10L 13/10: Prosody rules derived from text; Stress or intonation
- G10L 15/02: Feature extraction for speech recognition; Selection of recognition unit
- G10L 2013/105: {Duration}
- G10L 2015/025: {Phonemes, fenemes or fenones being the recognition units}
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING

## Abstract

A method for producing synthesized speech digital audio content, wherein: a feature extractor module receives an audio recording of a speaker's voice, extracts a plurality of acoustic features and converts them to an audio latent representation matrix; a phonemizing module receives as input a target text and converts the target text to a sequence of phonemes; a tokenizing module receives as input the sequence of phonemes of the target text; a linguistic encoder module receives as input the sequence of phoneme vectors and converts the sequence of phoneme vectors to a sequence of respective linguistic latent vectors; an acoustic model module produces a predicted audio latent representation matrix; and a vocoder module decodes the predicted audio latent representation matrix into the corresponding audio signal of the speech of the synthesized virtual voice.

## Claims

### Claim at position 1

1-13. (canceled)

### Claim 14

A method for producing synthesized speech digital audio content, comprising steps wherein:
a feature extractor module receives as input an audio recording of a speaker's voice, extracts a plurality of acoustic features from said audio recording, and converts said acoustic features to an audio latent representation matrix;
a phonemizing module of a text preprocessing module receives as input a target text and converts said target text to a sequence of phonemes;
a tokenizing module of said text preprocessing module receives as input said sequence of phonemes of said target text and converts said sequence of phonemes to a sequence of respective vectors of the phonemes of said target text, wherein each vector comprises a plurality of IDs that define a plurality of respective linguistic features of the phoneme of said target text;
a linguistic encoder module receives as input said sequence of phoneme vectors of said target text and converts said sequence of phoneme vectors to a sequence of respective linguistic latent vectors, wherein each vector represents a set of independent latent spaces;
an emotion predictor module of a speech emotion and emission recognition module receives as input said audio latent representation matrix, predicts an emotional state of the speech of a synthesized virtual voice, and produces as output a plurality of emotion signals in the time domain;
an emission predictor module of said speech emotion and emission recognition module receives as input said audio latent representation matrix, predicts an emission intensity of the speech of said synthesized virtual voice, and produces as output a plurality of emission signals in the time domain;
an acoustic model module receives as input said sequence of linguistic latent vectors and said plurality of emotion and emission signals in the time domain, predicts a latent representation of an audio signal of the speech of said synthesized virtual voice, and produces as output a predicted audio latent representation matrix; and
a vocoder module receives as input said predicted audio latent representation matrix and decodes said predicted audio latent representation matrix into said audio signal of the speech of said synthesized virtual voice.

### Claim 15 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
a voice space conversion module of a voice control module receives as input said audio latent representation matrix, and passes from a discrete voice space, defined by said audio latent representation matrix, to a continuous voice space, related to said audio latent representation matrix; and
a voice space mapping module of said voice control module receives as input said continuous voice space, related to said audio latent representation matrix, and creates a voice latent representation vector, which representing the voice timbre of said synthesized virtual voice:
wherein said acoustic model module further receives as input said voice latent representation vector.

### Claim 16 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
a duration predictor module of a duration control module receives as input said audio latent representation matrix and said sequence of linguistic latent vectors, predicts respective durations of the phonemes of said linguistic latent vectors, and produces as output a sequence of phoneme durations of said linguistic latent vectors;
wherein said acoustic model module further receives as input said sequence of phoneme durations.

### Claim 17 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
a pitch predictor module of a signal control module receives as input said sequence of linguistic latent vectors, said plurality of emotion and emission signals, optionally said voice latent representation vector, and optionally said sequence of phoneme durations, predicts a pitch for each frame of said audio latent representation, and produces as output a plurality of pitch signals in the time domain;
wherein said acoustic model module further receives as input said plurality of pitch signals.

### Claim 18 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
an energy predictor module of a signal control module receives as input said sequence of linguistic latent vectors, said plurality of emotion and emission signals, optionally said voice latent representation vector, and optionally said sequence of phoneme durations, predicts a magnitude for each frame of said audio latent representation, and produces as output a plurality of energy signals in the time domain;
wherein said acoustic model module further receives as input said plurality of energy signals.

### Claim 19 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
a trimming module of an audio preprocessing module receives as input said audio recording and removes portions of silence at the ends of said audio recording.

### Claim 20 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
a resampling module of an audio preprocessing module receives as input said audio recording and resamples said audio recording at a predetermined frequency.

### Claim 21 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
a loudness normalization module of an audio preprocessing module receives as input said audio recording and normalizes the loudness of said audio recording to a predefined value.

### Claim 22 (depends on 14)

The method for producing synthesized speech digital audio content according to claim 14, further comprising the step wherein:
a cleaning module of said text preprocessing module receives as input said target text and corrects typos present in said target text.

### Claim 23 (depends on 14)

A synthesized speech digital audio content, obtainable by means of a method for producing synthesized speech digital audio content according to claim 14.

### Claim 24 (depends on 14)

A system for producing synthesized speech digital audio content comprising modules configured to perform the steps of the method for producing synthesized speech digital audio content according to claim 14.
