Get the newsletter
Nvidia · AI / ML · Filed May 15, 2025 · USPTO published Nov 19, 2026 · This week's release

Nvidia Patents One Transcription System for Phone Calls and Studio Audio

Most speech-to-text tools are built for one kind of recording and stumble on the rest. Nvidia's patent application describes a single model that adapts to how each clip was recorded, from tinny phone calls to studio-quality audio.

Microphones and computers capturing spoken words and turning them into written text on local devices or through remote web services. Drawing from patent filing US 2026/0369278 A1.
Microphones and computers capturing spoken words and turning them into written text on local devices or through remote web services.All 26 drawings ↓
Publication number
US 2026/0369278 A1(US20260369278A1)
Applicant
Nvidia Corporation
Filed · Published
May 15, 2025 · Nov 19, 2026
Inventors
Harishchandra Dubey, Myungjong Kim, Oluwatobi Olabiyi
Status when we published
Waiting for an examiner
Document
20 claims
US classification
704/232
Examiner
FANG-WU, JOHN HONG (Art Unit 2653)

What Nvidia's one-model speech-to-text does

Every time a phone call, a podcast and a meeting recording land in the same transcription service, they arrive in very different shapes. A phone call captures a thin slice of sound detail, while a studio recording captures far more. Most systems are built for one kind of audio. They either convert everything else to fit, which can throw away detail, or they keep a separate model for each kind.

Nvidia's patent application describes one package instead. It holds several front ends tuned for different recording qualities, plus a single shared piece that writes out the words. The system checks how your clip was recorded and sends it to the matching front end, so the lossy conversion step is skipped.

The filing says the practical payoff is one model to store, update and retrain instead of a pile of near-duplicates.

From the filing · CLAIM 1
select, based at least on one or more properties of an input speech signal, at least one of an encoder from the plurality of encoders or a subset of audio channels from the set of audio channels; and generate an output corresponding to the input speech signal using at least the selected encoder or the selected subset of audio channels.

Translation: The system looks at the audio input and picks the best tools to process it into text.

How Nvidia picks the right encoder per clip

The system has several encoders (the parts of a speech model that listen to the audio and boil it down to an internal numerical summary) and one shared decoder (the part that turns that summary into words). Each encoder is tied to a different sampling rate, meaning how many times per second a recording measures the sound. The filing lists examples from 8 kHz phone audio up to 192 kHz professional recordings.

When a clip arrives, the system reads its properties, mainly the sampling rate and the number of audio channels (mono, stereo, surround), and routes it to the matching encoder. The audio is first chopped into short overlapping windows (one example uses 20 millisecond windows that overlap by 10), then converted into a spectral representation (a picture of which pitches are present over time). The model is built for a fixed number of channels, such as 16, and unused ones are set to zero. During training, empty channels can instead hold copies of real audio with added noise or echo.

Training is where the pieces get tied together. Clips of different qualities pass through their own encoders, and a joint fusion network combines the results before the shared decoder writes text. The errors from each path are combined as a weighted sum, with the weights adding up to one. At use time only one encoder runs, so the fusion step is skipped. The same setup can also be pointed at translation, such as German speech into English text.

From the filing · THE ABSTRACT
Approaches presented herein may be used to generate a text output of an input speech signal using an encoder from a plurality of encoders and a joint decoder associated with the plurality of encoders.

Translation: Nvidia's patented system converts spoken words into text by matching the audio with the right combination of processors.

Why one model for all audio matters to you

The filing's pitch is mostly about the companies that run speech services. Instead of keeping a separate model for phone audio, another for wideband audio, another for stereo, they keep one. The description says the shared decoder can save roughly 20 to 30 percent of the storage a model needs, and that retraining happens once instead of many times.

For you, the benefit shows up in accuracy. The filing says converting audio to fit a model can cause mismatches and more word errors, so skipping that step could mean cleaner transcripts of calls, meetings and voice commands. The description names video conferencing, in-car assistants and kiosks as places this could run. Remember this is a patent application, so it shows what Nvidia is describing, not a product you can use.

This is the 29th voice and speech AI filing from Nvidia we've tracked since May, adding to work like isolating one voice in a crowd and fixing spoken podcast errors.

Source: US 2026/0369278 A1 at the USPTO (PDF). Full patent text and figures are in the official publication.

Get our take in your Top Stories

Liked this breakdown? Add Patentlyze as a preferred source on Google, and our plain-English take shows up more often in your Top Stories the next time Nvidia patent news breaks.

There are more where this came from

We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.

The drawings · 26 sheets

26 drawing sheets from US 2026/0369278 A1 · click any drawing to enlarge

Three ways speech is turned into written words, drawing sheet 1 from US 2026/0369278 A1
Sheet 1 Three ways speech is turned into written words
Converting audio quality and channels before speech recognition, drawing sheet 2 from US 2026/0369278 A1
Sheet 2 Converting audio quality and channels before speech recognition
Feeding multiple audio tracks into a shared speech decoder, drawing sheet 3 from US 2026/0369278 A1
Sheet 3 Feeding multiple audio tracks into a shared speech decoder
Using multiple audio tracks to generate translated text, drawing sheet 4 from US 2026/0369278 A1
Sheet 4 Using multiple audio tracks to generate translated text
Choosing the right audio reader based on sample rates, drawing sheet 5 from US 2026/0369278 A1
Sheet 5 Choosing the right audio reader based on sample rates
Merging signals from different audio readers into one decoder, drawing sheet 6 from US 2026/0369278 A1
Sheet 6 Merging signals from different audio readers into one decoder
How training errors update the entire speech recognition model, drawing sheet 7 from US 2026/0369278 A1
Sheet 7 How training errors update the entire speech recognition model
Running one selected audio reader to transcribe or translate speech, drawing sheet 8 from US 2026/0369278 A1
Sheet 8 Running one selected audio reader to transcribe or translate speech
Show the other 18 sheets
Steps for turning speech into text using matching AI parts, drawing sheet 9 from US 2026/0369278 A1
Sheet 9 Steps for turning speech into text using matching AI parts
Training the AI model on different audio qualities, drawing sheet 10 from US 2026/0369278 A1
Sheet 10 Training the AI model on different audio qualities
Converting sound frequencies into written words, drawing sheet 11 from US 2026/0369278 A1
Sheet 11 Converting sound frequencies into written words
Connecting phones and cloud servers to run speech tools, drawing sheet 12 from US 2026/0369278 A1
Sheet 12 Connecting phones and cloud servers to run speech tools
Chip layout for calculating and storing AI voice data, drawing sheet 13 from US 2026/0369278 A1
Sheet 13 Chip layout for calculating and storing AI voice data
Dedicated processor blocks paired directly with memory, drawing sheet 14 from US 2026/0369278 A1
Sheet 14 Dedicated processor blocks paired directly with memory
Server center setup for running large voice models, drawing sheet 15 from US 2026/0369278 A1
Sheet 15 Server center setup for running large voice models
Computer parts that coordinate audio input and processing, drawing sheet 16 from US 2026/0369278 A1
Sheet 16 Computer parts that coordinate audio input and processing
Hardware parts inside a laptop or phone running speech tools, drawing sheet 17 from US 2026/0369278 A1
Sheet 17 Hardware parts inside a laptop or phone running speech tools
Connecting processors and sensors on a personal computer, drawing sheet 18 from US 2026/0369278 A1
Sheet 18 Connecting processors and sensors on a personal computer
Multi-core chip layout with graphics and memory sharing, drawing sheet 19 from US 2026/0369278 A1
Sheet 19 Multi-core chip layout with graphics and memory sharing
Preparing and training medical scan models for hospitals, drawing sheet 20 from US 2026/0369278 A1
Sheet 20 Preparing and training medical scan models for hospitals
Software stack linking AI models to graphics chips and servers, drawing sheet 21 from US 2026/0369278 A1
Sheet 21 Software stack linking AI models to graphics chips and servers
Human reviewers helping an AI model improve its labels, drawing sheet 22 from US 2026/0369278 A1
Sheet 22 Human reviewers helping an AI model improve its labels
Pre-trained models labeling raw images for further training, drawing sheet 23 from US 2026/0369278 A1
Sheet 23 Pre-trained models labeling raw images for further training
Turning user prompts into text responses using a language model, drawing sheet 24 from US 2026/0369278 A1
Sheet 24 Turning user prompts into text responses using a language model
An AI model reading a question to generate answers, drawing sheet 25 from US 2026/0369278 A1
Sheet 25 An AI model reading a question to generate answers
Predicting words sequentially using only a decoding model, drawing sheet 26 from US 2026/0369278 A1
Sheet 26 Predicting words sequentially using only a decoding model
Patent filing page

Reader comments

No comments yet

Start the discussion

Real name or a handle, either is fine. Comments are read by a person before they appear, so allow a little time. Keep it about the filing.