Emphasis Control for Parallel Neural TTS

March 16, 2023

Emphasis Control for Parallel Neural TTS

State-of-the-art neural text-to-speech (TTS) synthesis methods are able to generate speech with high fidelity while maintaining high-performance levels. However, the input to the TTS system does not often contain enough information in order to reconstruct the desired speech signal with the intended meaning and nuances. Therefore, TTS systems often generate average prosody learned from the database. This is not beneficial when there is a need to convey content in a specific style or emotion or highlight the part of the sentence that contributes new, non-derivable, or contrastive information. 

What is TTS?

TTS stands for Text-to-Speech, which refers to the process of converting written text into speech using computer algorithms. TTS technology allows computers and other devices to generate speech using a speech synthesis engine from written text, enabling them to read text aloud or communicate information to users who are persons with visual impairment, or learning disabilities, or individuals who prefer to listen to content in form of videos and audio

The engine uses rules and patterns to create phonemes, the basic units of speech and then combines them to produce words and sentences. This technology is used in a variety of applications, including assistive technology for persons with disabilities, language learning software, digital assistants, and voice-enabled devices such as smart speakers and virtual assistants.

What is Neural TTS?

Neural TTS or Neural Text-to-Speech, is a type of text-to-speech synthesis technology that enables neural networks to generate naturalistic speech from text. It is a branch of artificial intelligence, that aims to produce speech similar to human speech in terms of intonation, rhythm, and tone.

Traditional TTS systems utilized rule-based methods to generate speech, which requires the manual creation of linguistic rules, acoustic models, and other components. These systems often produced unnatural speech that lacked expressiveness.

The baseline model architecture is based on the FastSpeech parallel neural TTS model that converts an input phoneme sequence to a Mel spectrogram. The model consists of an encoder that converts the phoneme sequence to phoneme encodings. The encoder consists of an embedding layer that converts the phoneme sequence to phoneme embeddings. 

Afterward, a series of feed-forward transformer (FFT) intakes of the phoneme embeddings convert it into positional encodings. Each FFT consists of a self-attention layer and 1-D convolution layers along with layer normalization and dropout. 

The phoneme encodings are then fed to the feature predictors that predict the phoneme-wise pitch, duration, and energy. The feature predictors consist of 1-D convolution layers, layer normalization, and dropout similar to the variance adaptors utilized. The predicted phoneme-wise pitch and energy features are quantized, passed through an embedding layer, and added to the phoneme encodings, to form the decoder input. All the feature predictors are trained using teacher forcing. The decoder inputs are upsampled based on the predicted phoneme-wise durations. The FastSpeech decoder which consists of FFT blocks is replaced here with a series of dilated convolution stacks, which improves model inference speed. This model achieves 150× faster than real-time synthesis on a GPU and 100× faster on a mobile device. 

For example, to analyze how the models with different emphasis features respond to varying degrees of emphasis control, we synthesized 192 test sentences for the VarianceBased, WaveletBased, and Combined models and each voice using varying degrees of emphasis modification. This was done by adding different bias values to the predicted emphasis features for the phonemes of the emphasized word. Larger values of this bias should result in increased emphasis while negative values should correspond to de-emphasis. For the VarianceBased method with 2 emphasis features, both features were modified by the same amount. The emphasis features influence the pitch, energy, and phoneme duration of the models, and therefore, we do measure the resulting changes in these three features before and after the changes in the emphasis features to assess the emphasis control. This shows the change in the mean and standard deviations for pitch, energy, and phoneme duration after applying emphasis averaged per phone in the emphasized words.

What is Emphasis Control for Parallel Neural TTS?

Emphasis control is a technique used in text-to-speech (TTS) systems to adjust the prosodic features of synthesized speech, such as pitch, duration, and loudness, to convey certain emotions or emphasis. In the context of parallel neural TTS,  neural networks are used to generate speech from text, emphasis control can be achieved by conditioning the neural network on additional input features related to emphasis, such as the desired pitch or duration of certain words.

One approach to emphasis control in parallel neural TTS is to use a multi-speaker TTS model, where each speaker is associated with a different emphasis style. This allows the system to synthesize speech with different emphasis styles by simply switching the speaker input. Another approach is to use an attention mechanism to selectively amplify or attenuate certain input features, such as the pitch or duration of certain words, based on a user-defined emphasis signal. This can be achieved by incorporating an attention module into the neural network architecture and training it to attend to the relevant input features based on the emphasis signal.

In addition to these approaches, recent research has also explored the use of adversarial training to learn emphasis styles from a small amount of labeled data, as well as the use of reinforcement learning to optimize the emphasis control policy based on user feedback. Overall, emphasis control is an important technique for improving the expression and naturalness of synthesized speech in parallel neural TTS systems and is an active area of research in the field.

History of TTS

Text-to-speech (TTS) technology has been around for several decades and has undergone significant advancements and changes. One of the earliest developments of TTS can be traced back to the Bell Labs in the 1930s, when a system was developed that could produce speech sounds by generating electrical signals. However, it was not until the 1960s that computer-based TTS systems came into existence

In the early days of TTS, systems used rule-based methods to produce speech sounds. These systems used a set of rules and linguistic knowledge to determine the pronunciation of words and generate the corresponding sounds. However, these systems were limited in their ability to produce humanistic speech.

In the 1980s, a new approach to TTS known as the "concatenative" method was developed. This approach involved recording individual sounds or phonemes and then combining them to produce speech. This approach allowed for more humanistic speech and was used in many TTS systems throughout the 1990s and 2000s.

Recent advancements in machine learning and artificial intelligence have led to the development of ‘neural’ TTS systems. These systems use deep learning algorithms to generate speech, allowing for even more natural-sounding and expressive speech. TTS technology is used in a wide range of applications, from accessibility tools for persons with visual impairments or virtual assistance through chatbots. 

Datasets used to deploy ML models for Parallel Neural TTS

These are a few of the several datasets that have been used for training parallel neural TTS systems: 

  1. LJSpeech: A dataset of approximately 13,100 short audio clips of a single speaker reading passages from 7 non-fiction books, paired with their corresponding transcripts. This dataset is widely used in the TTS community and is freely available for download.

Github source: https://github.com/keithito/tacotron/commit/67e16fcbf570852a828909633f22d5bbdea8ac19

  1. LibriTTS: A dataset of over 5,000 hours of reading English speech from audiobooks, with accompanying text. This dataset is designed for TTS research and is available for download from the Mozilla Common Voice website.

Github source: https://github.com/tensorflow/datasets/blob/master/docs/catalog/libritts.md

  1. M-AILABS Speech Dataset: A dataset of approximately 3,000 hours of high-quality speech recordings from a single female speaker reading English sentences, paired with their corresponding text. This dataset is also freely available for download.

Github source: https://github.com/MycroftAI/mimic2/blob/master/datasets/mailabs.py

  1. Blizzard Challenge: A dataset of speech recordings and text transcriptions from multiple speakers in different languages. This dataset is commonly used for TTS research and is available for download from the Blizzard Challenge website.

Github source: https://github.com/Tomiinek/Blizzard2013_Segmentation

  1. Common Voice: A dataset of crowd-sourced speech recordings in multiple languages, paired with their corresponding text transcriptions. This dataset is freely available for download from the Mozilla Common Voice website.

Github source: https://github.com/common-voice/common-voice-bundler

These datasets have been used in various TTS systems such as Tacotron, Transformer TTS, FastSpeech, and more.

How does NVIDIA A100 help in parallel neural TTS?

The NVIDIA A100 GPU is a powerful computing resource that can help improve the performance and efficiency of parallel neural TTS systems, enabling faster training times and higher-quality speech synthesis in several ways:

  1. Increased Performance: The A100 GPU is designed specifically for high-performance computing, including deep learning workloads. It provides up to 20 times greater computing power than the previous generation of NVIDIA GPUs, which allows for faster training and inference times.
  2. Improved Efficiency: The A100 GPU uses NVIDIA's latest Tensor Cores technology, which is designed to accelerate matrix math operations commonly used in deep learning workloads. This results in improved efficiency and faster training times.
  3. Large Memory Capacity: The A100 GPU features up to 80 GB of high-bandwidth memory, which allows for larger neural networks to be trained and stored in memory, improving the overall performance of the parallel TTS system.

Launch A100 80GB Cloud GPU on E2E Cloud for training a TTS model

  • Go to Compute> GPU> NVIDIA- A100 80GB.
  • Click on “Create” and choose your plan.

  • Choose your required security, backup, and network settings and click on “Create My Node”.

  • The launched plan will appear in your dashboard once it starts running.

After launching the A100 80GB Cloud GPU from the Myaccount portal, you can deploy any TTS  model for emphasis control. 

E2E Networks is the leading accelerated Cloud Computing player which provides the latest Cloud GPUs at a great value. Connect with us at sales@e2enetworks.com

Request a free trial here 

Latest Blogs
This is a decorative image for: A Complete Guide To Customer Acquisition For Startups
October 18, 2022

A Complete Guide To Customer Acquisition For Startups

Any business is enlivened by its customers. Therefore, a strategy to constantly bring in new clients is an ongoing requirement. In this regard, having a proper customer acquisition strategy can be of great importance.

So, if you are just starting your business, or planning to expand it, read on to learn more about this concept.

The problem with customer acquisition

As an organization, when working in a diverse and competitive market like India, you need to have a well-defined customer acquisition strategy to attain success. However, this is where most startups struggle. Now, you may have a great product or service, but if you are not in the right place targeting the right demographic, you are not likely to get the results you want.

To resolve this, typically, companies invest, but if that is not channelized properly, it will be futile.

So, the best way out of this dilemma is to have a clear customer acquisition strategy in place.

How can you create the ideal customer acquisition strategy for your business?

  • Define what your goals are

You need to define your goals so that you can meet the revenue expectations you have for the current fiscal year. You need to find a value for the metrics –

  • MRR – Monthly recurring revenue, which tells you all the income that can be generated from all your income channels.
  • CLV – Customer lifetime value tells you how much a customer is willing to spend on your business during your mutual relationship duration.  
  • CAC – Customer acquisition costs, which tells how much your organization needs to spend to acquire customers constantly.
  • Churn rate – It tells you the rate at which customers stop doing business.

All these metrics tell you how well you will be able to grow your business and revenue.

  • Identify your ideal customers

You need to understand who your current customers are and who your target customers are. Once you are aware of your customer base, you can focus your energies in that direction and get the maximum sale of your products or services. You can also understand what your customers require through various analytics and markers and address them to leverage your products/services towards them.

  • Choose your channels for customer acquisition

How will you acquire customers who will eventually tell at what scale and at what rate you need to expand your business? You could market and sell your products on social media channels like Instagram, Facebook and YouTube, or invest in paid marketing like Google Ads. You need to develop a unique strategy for each of these channels. 

  • Communicate with your customers

If you know exactly what your customers have in mind, then you will be able to develop your customer strategy with a clear perspective in mind. You can do it through surveys or customer opinion forms, email contact forms, blog posts and social media posts. After that, you just need to measure the analytics, clearly understand the insights, and improve your strategy accordingly.

Combining these strategies with your long-term business plan will bring results. However, there will be challenges on the way, where you need to adapt as per the requirements to make the most of it. At the same time, introducing new technologies like AI and ML can also solve such issues easily. To learn more about the use of AI and ML and how they are transforming businesses, keep referring to the blog section of E2E Networks.

Reference Links




This is a decorative image for: Constructing 3D objects through Deep Learning
October 18, 2022

Image-based 3D Object Reconstruction State-of-the-Art and trends in the Deep Learning Era

3D reconstruction is one of the most complex issues of deep learning systems. There have been multiple types of research in this field, and almost everything has been tried on it — computer vision, computer graphics and machine learning, but to no avail. However, that has resulted in CNN or convolutional neural networks foraying into this field, which has yielded some success.

The Main Objective of the 3D Object Reconstruction

Developing this deep learning technology aims to infer the shape of 3D objects from 2D images. So, to conduct the experiment, you need the following:

  • Highly calibrated cameras that take a photograph of the image from various angles.
  • Large training datasets can predict the geometry of the object whose 3D image reconstruction needs to be done. These datasets can be collected from a database of images, or they can be collected and sampled from a video.

By using the apparatus and datasets, you will be able to proceed with the 3D reconstruction from 2D datasets.

State-of-the-art Technology Used by the Datasets for the Reconstruction of 3D Objects

The technology used for this purpose needs to stick to the following parameters:

  • Input

Training with the help of one or multiple RGB images, where the segmentation of the 3D ground truth needs to be done. It could be one image, multiple images or even a video stream.

The testing will also be done on the same parameters, which will also help to create a uniform, cluttered background, or both.

  • Output

The volumetric output will be done in both high and low resolution, and the surface output will be generated through parameterisation, template deformation and point cloud. Moreover, the direct and intermediate outputs will be calculated this way.

  • Network architecture used

The architecture used in training is 3D-VAE-GAN, which has an encoder and a decoder, with TL-Net and conditional GAN. At the same time, the testing architecture is 3D-VAE, which has an encoder and a decoder.

  • Training used

The degree of supervision used in 2D vs 3D supervision, weak supervision along with loss functions have to be included in this system. The training procedure is adversarial training with joint 2D and 3D embeddings. Also, the network architecture is extremely important for the speed and processing quality of the output images.

  • Practical applications and use cases

Volumetric representations and surface representations can do the reconstruction. Powerful computer systems need to be used for reconstruction.

Given below are some of the places where 3D Object Reconstruction Deep Learning Systems are used:

  • 3D reconstruction technology can be used in the Police Department for drawing the faces of criminals whose images have been procured from a crime site where their faces are not completely revealed.
  • It can be used for re-modelling ruins at ancient architectural sites. The rubble or the debris stubs of structures can be used to recreate the entire building structure and get an idea of how it looked in the past.
  • They can be used in plastic surgery where the organs, face, limbs or any other portion of the body has been damaged and needs to be rebuilt.
  • It can be used in airport security, where concealed shapes can be used for guessing whether a person is armed or is carrying explosives or not.
  • It can also help in completing DNA sequences.

So, if you are planning to implement this technology, then you can rent the required infrastructure from E2E Networks and avoid investing in it. And if you plan to learn more about such topics, then keep a tab on the blog section of the website

Reference Links



This is a decorative image for: Comprehensive Guide to Deep Q-Learning for Data Science Enthusiasts
October 18, 2022

A Comprehensive Guide To Deep Q-Learning For Data Science Enthusiasts

For all data science enthusiasts who would love to dig deep, we have composed a write-up about Q-Learning specifically for you all. Deep Q-Learning and Reinforcement learning (RL) are extremely popular these days. These two data science methodologies use Python libraries like TensorFlow 2 and openAI’s Gym environment.

So, read on to know more.

What is Deep Q-Learning?

Deep Q-Learning utilizes the principles of Q-learning, but instead of using the Q-table, it uses the neural network. The algorithm of deep Q-Learning uses the states as input and the optimal Q-value of every action possible as the output. The agent gathers and stores all the previous experiences in the memory of the trained tuple in the following order:

State> Next state> Action> Reward

The neural network training stability increases using a random batch of previous data by using the experience replay. Experience replay also means the previous experiences stocking, and the target network uses it for training and calculation of the Q-network and the predicted Q-Value. This neural network uses openAI Gym, which is provided by taxi-v3 environments.

Now, any understanding of Deep Q-Learning   is incomplete without talking about Reinforcement Learning.

What is Reinforcement Learning?

Reinforcement is a subsection of ML. This part of ML is related to the action in which an environmental agent participates in a reward-based system and uses Reinforcement Learning to maximize the rewards. Reinforcement Learning is a different technique from unsupervised learning or supervised learning because it does not require a supervised input/output pair. The number of corrections is also less, so it is a highly efficient technique.

Now, the understanding of reinforcement learning is incomplete without knowing about Markov Decision Process (MDP). MDP is involved with each state that has been presented in the results of the environment, derived from the state previously there. The information which composes both states is gathered and transferred to the decision process. The task of the chosen agent is to maximize the awards. The MDP optimizes the actions and helps construct the optimal policy.

For developing the MDP, you need to follow the Q-Learning Algorithm, which is an extremely important part of data science and machine learning.

What is Q-Learning Algorithm?

The process of Q-Learning is important for understanding the data from scratch. It involves defining the parameters, choosing the actions from the current state and also choosing the actions from the previous state and then developing a Q-table for maximizing the results or output rewards.

The 4 steps that are involved in Q-Learning:

  1. Initializing parameters – The RL (reinforcement learning) model learns the set of actions that the agent requires in the state, environment and time.
  2. Identifying current state – The model stores the prior records for optimal action definition for maximizing the results. For acting in the present state, the state needs to be identified and perform an action combination for it.
  3. Choosing the optimal action set and gaining the relevant experience – A Q-table is generated from the data with a set of specific states and actions, and the weight of this data is calculated for updating the Q-Table to the following step.
  4. Updating Q-table rewards and next state determination – After the relevant experience is gained and agents start getting environmental records. The reward amplitude helps to present the subsequent step.  

In case the Q-table size is huge, then the generation of the model is a time-consuming process. This situation requires Deep Q-learning.

Hopefully, this write-up has provided an outline of Deep Q-Learning and its related concepts. If you wish to learn more about such topics, then keep a tab on the blog section of the E2E Networks website.

Reference Links



This is a decorative image for: GAUDI: A Neural Architect for Immersive 3D Scene Generation
October 13, 2022

GAUDI: A Neural Architect for Immersive 3D Scene Generation

The evolution of artificial intelligence in the past decade has been staggering, and now the focus is shifting towards AI and ML systems to understand and generate 3D spaces. As a result, there has been extensive research on manipulating 3D generative models. In this regard, Apple’s AI and ML scientists have developed GAUDI, a method specifically for this job.

An introduction to GAUDI

The GAUDI 3D immersive technique founders named it after the famous architect Antoni Gaudi. This AI model takes the help of a camera pose decoder, which enables it to guess the possible camera angles of a scene. Hence, the decoder then makes it possible to predict the 3D canvas from almost every angle.

What does GAUDI do?

GAUDI can perform multiple functions –

  • The extensions of these generative models have a tremendous effect on ML and computer vision. Pragmatically, such models are highly useful. They are applied in model-based reinforcement learning and planning world models, SLAM is s, or 3D content creation.
  • Generative modelling for 3D objects has been used for generating scenes using graf, pigan, and gsn, which incorporate a GAN (Generative Adversarial Network). The generator codes radiance fields exclusively. Using the 3D space in the scene along with the camera pose generates the 3D image from that point. This point has a density scalar and RGB value for that specific point in 3D space. This can be done from a 2D camera view. It does this by imposing 3D datasets on those 2D shots. It isolates various objects and scenes and combines them to render a new scene altogether.
  • GAUDI also removes GANs pathologies like mode collapse and improved GAN.
  • GAUDI also uses this to train data on a canonical coordinate system. You can compare it by looking at the trajectory of the scenes.

How is GAUDI applied to the content?

The steps of application for GAUDI have been given below:

  • Each trajectory is created, which consists of a sequence of posed images (These images are from a 3D scene) encoded into a latent representation. This representation which has a radiance field or what we refer to as the 3D scene and the camera path is created in a disentangled way. The results are interpreted as free parameters. The problem is optimized by and formulation of a reconstruction objective.
  • This simple training process is then scaled to trajectories, thousands of them creating a large number of views. The model samples the radiance fields totally from the previous distribution that the model has learned.
  • The scenes are thus synthesized by interpolation within the hidden space.
  • The scaling of 3D scenes generates many scenes that contain thousands of images. During training, there is no issue related to canonical orientation or mode collapse.
  • A novel de-noising optimization technique is used to find hidden representations that collaborate in modelling the camera poses and the radiance field to create multiple datasets with state-of-the-art performance in generating 3D scenes by building a setup that uses images and text.

To conclude, GAUDI has more capabilities and can also be used for sampling various images and video datasets. Furthermore, this will make a foray into AR (augmented reality) and VR (virtual reality). With GAUDI in hand, the sky is only the limit in the field of media creation. So, if you enjoy reading about the latest development in the field of AI and ML, then keep a tab on the blog section of the E2E Networks website.

Reference Links




Build on the most powerful infrastructure cloud

A vector illustration of a tech city using latest cloud technologies & infrastructure