TTS, a ubiquitous technology in modern applications, leverages artificial intelligence and AI learning to transform vast audio datasets into invaluable digital assistants. The initial step in creating a TTS voice database involves defining its specific purpose and characteristics, which subsequently determine the recording specifications and methodology.

Familiar examples of TTS applications include smartphone voice assistants like Siri and Google Assistant. Our team at MuScene (Audio Recording and Production Department & Voice Forensics Lab) has been involved in various projects such as: transportation voice navigation systems, international smartphone voice assistant systems, interactive home robot question-answer systems, linguistic/academic voice database sampling, electroglottography (EGG), and voice ID sample collection and research. Essentially, TTS technology and its AI engine can be integrated into a wide range of interactive products, breathing new life into otherwise inanimate devices and enhancing user convenience. Moreover, TTS can create innovative business models and applications for enterprises.

If your team is embarking on a new TTS project but lacks the necessary experience, we strongly recommend consulting with our team of experts. We can provide you with a comprehensive overview of TTS voice sampling (including electroglottography) and help you assess the resources required for your project.
With thousands of recording studios and music workshops in Taiwan alone, choosing the right one for your TTS project can be overwhelming.

Based on the insights from our Voice Forensics Lab at MuScene, we've identified several critical parameters, roles, and considerations for TTS projects. These include:
• Studio acoustics: low-frequency standing waves, background noise, sound insulation, RT60, etc.
• Professional audio engineers
• Recording assistants
• Professional voice actors
• Linguists
• Project managers
• Audio interface’s ADDA, THD Value, Pre-Amp SN Ratio
• Microphone frequency response and sensitivity
• Cable total harmonic distortion

To meet the rigorous TTS sampling standards set by international corporations, a recording studio must essentially operate at the level of a scientific acoustics laboratory. Traditional commercial recording studios often fall short of these requirements.

The Heart of TTS: The Voice Actor
The voice actor is undoubtedly the soul of a TTS project, as their voice is the most memorable aspect of the final product. Therefore, their voice must align with the product or brand's positioning, image, and quality. Beyond a pleasant voice, linguistic factors such as regional accent, language proficiency, tone, articulation, consistency, listenability, and electroglottography (EGG) noise should also be considered. Ideally, the voice actor should convey a warm and sincere tone to facilitate a smooth project execution.

The Importance of Linguistic Expertise
Our linguists work closely with our voice actors to ensure that your TTS data is accurate, natural, and culturally appropriate. They play a critical role in:
• Text selection (Scrip Design) and preparation
• Defining pronunciation guidelines
• Ensuring linguistic consistency

By leveraging our linguistic expertise, you can create a TTS solution that is both effective and engaging.
The Role of the Project Manager in TTS

A TTS recording project involves a multitude of tasks, from text volume assessment and recording schedule planning to post-production editing, segmentation, annotation, AI engine integration, and final project closure. Given the complexity and interdisciplinary nature of such projects, a Project Manager is essential to oversee the entire process, ensuring effective communication, coordination, and troubleshooting.

For instance, while the quantity of a TTS dataset is often defined by the number of words or sentences, we typically measure it by the total recording time. A dataset of approximately 2,000 sentences (around 30,000 words) might require 25 hours of recording but yield only 3-4 hours of usable audio due to various factors such as content type and editing.

The Project Manager is responsible for coordinating the efforts of various team members, including linguists, voice actors, audio engineers, and post-production specialists, to ensure that the project stays on schedule and within budget.

Once the audio recordings are completed, a quality assurance team conducts a thorough review to identify errors, omissions, and areas that require additional recording. This iterative process continues until the dataset is comprehensive and accurate.

In the overall TTS project lifecycle, the recording phase represents the midpoint. It is preceded by extensive planning and research and followed by the development and testing of the TTS engine.

MuScene Voice Forensics Laboratory, with our state-of-the-art recording studio and voice forensics laboratory, offers a comprehensive suite of services for TTS project development. We are committed to providing our clients with the highest quality voice data and ensuring the success of their TTS applications.
Our team of experienced audio engineers and researchers has a deep understanding of the technical and linguistic aspects of TTS. We can help you with:
• Customized TTS project planning
• Script design and linguistic analysis
• Voice data collection and processing
By partnering with MuScene, you can benefit from:
• Reduced project risks
• Improved product quality
• Enhanced global competitiveness