How do you generate AI voiceovers for video ads?

How does Google Cloud Text-to-Speech work?

Google Cloud Text-to-Speech is an API that converts text into natural-sounding, synthetic human speech. This process allows you to create playable audio data files from written input. These audio files can then be used to augment media such as videos or audio recordings, provided they comply with the Google Cloud Platform Terms of Service and all applicable laws.

The service takes either plain text or Speech Synthesis Markup Language (SSML) as input. It then generates raw audio data as a base64-encoded string. Before this audio can be played by an application or integrated into a video, the base64-encoded string must be decoded into a standard audio file format, such as MP3, Linear16 (which is the encoding used in WAV files), or OGG Opus.

To begin using Google Cloud Text-to-Speech, you must first enable the service on a Google Cloud project and ensure that billing is enabled for it. If you are working from a local shell, you will also need to create local authentication credentials for your user account. Once these prerequisites are met, you can send a synthesize request to the Cloud Text-to-Speech API to create an audio file of synthetic human speech.

What types of voices are available?

Google Cloud Text-to-Speech offers a wide selection of voices, with over 380 options available across more than 75 languages and their variants, including Mandarin, Hindi, Spanish, Arabic, and Russian. The service generates audio with natural, human-like quality, making the speech sound like a real person. When you send a synthesis request, you must specify which voice will 'speak' the words.

Several types of voices are available, each with distinct characteristics. WaveNet-generated voices, for example, are often perceived as more warm and human-like than other synthetic voices because their models are trained using raw audio samples of actual humans speaking. Chirp 3 HD voices are driven by technology that captures nuances in human intonation, which can make conversations more engaging. However, Chirp 3 HD voices do not support SSML input, speaking rate and pitch-audio parameters, or the A-Law audio encoding.

Other voice options include Studio voices, which are specifically designed for news reading and broadcast content, and Neural2 voices, which allow users to leverage Custom Voice technology without needing to train their own custom voice. This variety enables you to precisely dictate style, accent, pace, tone, and emotional expression through simple natural-language prompts in over 75 locales, or even create a unique voice to represent your brand across all customer touchpoints.

How can you customize the voiceover?

You can customize the voiceover generated by Google Cloud Text-to-Speech in several ways to achieve the desired sound for your video ad. The service allows you to precisely dictate elements such as style, accent, pace, tone, and emotional expression. These adjustments can be controlled through simple natural-language prompts, which are supported in over 75 locales.

For more detailed control over pronunciation and timing, you can use Speech Synthesis Markup Language (SSML) tags. SSML lets you insert pauses, format numbers, dates, and times, and provide other specific pronunciation instructions. For example, you can use SSML to ensure acronyms are pronounced correctly or to add specific pauses for dramatic effect. Additionally, you can configure the speaking rate, pitch, volume, and sample rate hertz for the synthetic speech.

It is important to note that not all customization options are available for every voice type. For instance, Chirp 3 HD voices do not support SSML input, speaking rate, or pitch-audio parameters. When selecting a voice and planning your customizations, consider these limitations to ensure compatibility with your desired output.

What audio formats does it support?

Google Cloud Text-to-Speech supports converting text into several widely used audio formats. These include MP3, Linear16, and OGG Opus, among others. Linear16 is the encoding commonly used in WAV files, making it compatible with many audio editing and playback systems.

When the speech synthesis process generates audio, it initially produces raw audio data as a base64-encoded string. This string is not directly playable. To use the audio in your video ads or other applications, you must decode this base64-encoded string into one of the supported audio file formats. This decoding step transforms the raw data into a playable audio file that can be easily integrated into your media projects.

How is Google Cloud Text-to-Speech priced?

Google Cloud Text-to-Speech is priced based on the number of characters sent to the service each month to be synthesized into audio. This means your cost will depend on the total volume of text you convert into speech.

The service also offers a free tier for WaveNet voices. The first 1 million characters synthesized using WaveNet voices are free each month. This allows users to experiment with the service and generate a significant amount of audio without incurring charges, particularly for the warm and human-like WaveNet voices.

Common questions

What is Google Cloud Text-to-Speech?

Google Cloud Text-to-Speech is an API that converts text into natural-sounding, synthetic human speech. It allows you to create playable audio data files from written input, which can then be used to augment media like videos or audio recordings.

How many voices and languages are available?

The service offers a selection of over 380 voices across more than 75 languages and their variants. This includes major languages like Mandarin, Hindi, Spanish, Arabic, and Russian, providing extensive options for global content.

Can I customize the voice's emotional expression?

Yes, you can precisely dictate the style, accent, pace, tone, and emotional expression of the voice. This customization is steerable through simple natural-language prompts in over 75 locales, allowing for fine-tuned control over the voiceover's delivery.

What is SSML and how is it used?

SSML stands for Speech Synthesis Markup Language. It allows you to insert specific instructions into your text input, such as adding pauses, formatting numbers, dates, and times, and providing other pronunciation details. This helps to refine the audio data created by the service.

Are there any free options for using Google Cloud Text-to-Speech?

Yes, Google Cloud Text-to-Speech offers a free tier. The first 1 million characters synthesized using WaveNet voices are free each month. Pricing beyond this free tier is based on the total number of characters sent to the service for synthesis.