Prosody is the rhythm and intonation of language, the way that you speak. This includes how fast you speak, where you put more emphasis on the words, and the volume of certain words.
Prosody in sales calls is really important especially if you are using voice AI agents because it shapes how the message is perceived. If the tone feels flat and lacks emphasis, the pitch won’t hit the recipient and overall business performance falls. So, prosody portrays a business as a trustworthy and confident entity, thereby indirectly planting the seed of loyalty.
Emphasis, pauses and intonation are the main components of prosody as each of them highlights the important component of conversation. For example, emphasis during conversation highlights key benefits, pauses agents take signal turn-taking, and intonation clarifies whether the AI is asking a question or making a statement. Without proper prosody, important sales points are harder to understand and easy to miss.
If AI calling agents sound unnatural, flat , and robotic sales are directly affected therefore it is very important for businesses to tune prosody of AI sales calls. Modern text-to-speech (TTS) systems use neural networks to imitate natural speech rhythms. These systems analyze linguistic cues like sentence structure and punctuation to generate appropriate intonation, stress, and pauses. This is one of the ways for tuning AI prosody by using advanced technologies, let’s learn and understand the other ways to tune it!
Explore more: Technical Foundations Behind High-Performance AI Calling Systems
Ways of Tuning AI Prosody for Sales Calls
There are two main ways to tune prosody. One is to use advanced technologies, more developed than the ones available now. To use this method we obviously have to wait till new and highly advanced models appear in the tech field. The other is to optimize the existing setup to tune prosody. And that one is the most reasonable way and can be done by optimizing the features of already available models. Let’s learn how to tune prosody by keeping the underlying architecture the same!
1. Speech Synthesis Markup Language (SSML)
Speech Synthesis Markup Language (SSML) is an XML-based markup language that is supported by many text-to-speech models like Google, Amazon, and Microsoft. It allows developers to tune prosody by controlling the vocal characteristics of voice agents.
Common SSML tags can be used to improve text-to-speech recordings. These tags are
- break/pause
- phoneme
- prosody
- Say-as
- emphasis
But in this blog, we will focus only on prosody, break/pause tag, and emphasis tags because if they are utilized perfectly, they improve rhythm and intonation of language!
i. Prosody tag
The prosody tag is the main way through which developers can tune sales calls for prosody. The prosody tag has five main elements:
- Rate: It defines how fast or slow the AI agent speaks
- Pitch: It defines how high or low the sound of an agent is. It allows humans to perceive sounds as being “high” or “low.”
- Volume: it explains how loud the voice is delivered by the agent.
- Contour: It controls the melody of speech. It is used to show emotions, emphasis and questions.
- Range: It explains how wide the pitch variation is. If the pitch range is narrow, the voice will sound calm, controlled, or monotone. And if the pitch range is wide, the voice sounds expressive, energetic.
When these elements are tuned perfectly, they reduce interruptions, improve engagement, and make AI voice sound more human and natural.
ii. Break/Pause tag
Break/pause tag is used to insert the pauses between words, phrases, and sentences. Break elements include
- commas
- quotation marks
- periods
Each of these break elements has a default pause duration ranging from seconds to milliseconds. To keep it sound natural, always set that limit to less than 5000 milliseconds. Besides setting exact timings of pauses, developers can introduce pauses using strength elements like
- x-weak
- weak
- medium
- strong
- x-strong
Each of these elements corresponds to the strength of the pause from very short to very long pauses. While using break elements, you have to be cautious. Using them too often can disrupt the natural pauses and voice sounds unnatural. But small controlled pauses can make AI sound like its thinking and rather than reading. Moreover, every controlled pause helps listeners to process information more easily.
iii. Emphasis Tag
The emphasis tag is used to subtly stress important words or phrases. It modifies delivery by adjusting loudness, speed, or prominence so that key terms stand out. In sales calls, emphasis is effective when applied sparingly to value-driven words such as benefits, guarantees, or time relevance. But its overuse can make the voice sound scripted or promotional.
Together, these three tags; <prosody>, <break>, and <emphasis>; are the most practical ways to tune prosody if used carefully!
How Does SSML Help Tune Prosody in Sales Conversations?
Opening
The opening of sales calls are friendly and easygoing. At this point, use a slow to medium speaking rate and add a short pause after the name/brand using SSML tags like <break>, <prosody rate=…>.
Qualification
In qualification, keep the tone neutral to avoid sounding salesy. Keep pitch changes minimal and use question intonation. In SSML, avoid heavy emphasis and keep <prosody> adjustments modest. You can also adjust contour and range to keep delivery controlled, with short pause before key question.
Value pitch
In the value pitch, add a bit more energy but don’t make it advertisement-like. You can achieve this pattern by keeping speaking rate stable, using <emphasis> for unique selling points and a small increase in volume. Excessive emphasis can make speech sound promotional and salesy.
Objection handling
For object handling, calmness is key. You can achieve it by slowing speaking rate, reducing energy, and adding slightly longer pauses. These features make the agent sound patient and thoughtful. Use <break> to insert intentional silence, and <prosody rate=…> to slow the delivery.
Closing
Make the closing of your sales call confident by using a steady speaking rate. Moreover, add a short pause after asking for the follow-up time or date so the customer feels invited to answer. This can be done by <break> placement that improves conversational flow without changing the script.
2. Tuning Prosody Through Voice Settings
Many voice/TTS vendors let you tune prosody by giving you control over voice-level settings. By optimizing these settings, you can fine tune prosody without inserting tags into the text. They are focused on changing how the voice model delivers the words. By optimizing these settings you can control whether the uttered words are consistent, expressive, or animated. .
For example, ElevenLabs provides setting options like
- Stability
- Similarity
- Style.
Stability controls how consistent the voice is. Higher stability makes the voice more steady and predictable, while lower stability makes it more expressive but less controlled. Similarity controls how closely the voice follows its original sound. Higher similarity usually improves clarity but pushing it too high can cause issues. Style controls how emotional or animated the voice sounds.
Other platforms may not use sliders, but they still affect prosody through voice selection.
The practical approach to optimize the voice setting is:
- Choose the one baseline voice out of the available voices provided by the vendor. Select the one that aligns best with your brand’s voice. And choose one backup voice in case the main voice has technical issues, or does not work well for a specific customer group.
- Then, when you tune these settings (Stability/Similarity/Style). At a time, change only 1 or 2 setting options and test. This is because these controls interact with each other and multi-slider changes can make the voice sound totally different from your reference or outcome you thought in your mind.
3. Follow Copywriting rules that improve prosody automatically
One of the ways to improve prosody is using speech-first style to write scripts for sales calls.
Write in short clauses and short sentences for voice calls because long and nested sentences are harder to utter and sometimes result in wrong emphasis and awkward timing. Use complete sentence tag [<s>…</s>] for wrapping the full sentence. It highlights the start and end of a sentence. As a result, text-to-speech acts more naturally by pausing correctly at the end of the sentence, keeping the rhythm natural, and applying prosody changes smoothly.
Use punctuation for natural pauses instead of extra tags. TTS systems handle common punctuation naturally like pausing after periods and uttering sentences with question mark as question rather than simple statement. Overriding default pauses with manual breaks can backfire and make the speech feel unnatural.
Avoid adding “fake human” fillers like “um,” “uh,” or “you know” unless you need them. Moreover, increase readability of call scripts by expanding abbreviations, numbers, and symbols in a way people utter naturally. For example, convert $42.50 into forty-two dollars and fifty cents. This reduces reading mistakes and improves natural flow without needing prosody tags.
FAQs about Tuning Prosody for Sales Calls
1. What is prosody in AI sales calls and why does it matter?
Prosody is the rhythm and intonation of language, the way that you speak. This includes how fast you speak, where you put more emphasis on the words, and the volume of certain words. Prosody in sales calls is really important especially if you are using voice AI agents because it shapes how the message is perceived. If the tone feels flat and lacks emphasis, the pitch won’t hit the recipient and overall business performance falls. So, prosody portrays a business as a trustworthy and confident entity, thereby indirectly planting the seed of loyalty.
2. How does SSML help tune prosody for AI voice agents?
SSML allows developers to tune prosody without changing script and underlying architecture by using tags like <prosody>, <break>, and <emphasis>. With the help of these tags, different characteristics of speech can be improved that make AI voice sound more natural.
3. Which SSML tags are most important for improving sales call prosody?
The most important tag for tuning prosody is <prosody>. It allows you to control rhythm and intonation of language via five prosody elements. These elements are
- rate
- Pitch
- Volume
- Contour
- range
4. Can AI prosody be improved without using SSML?
Yes. Many voice platforms offer voice-level settings like stability, similarity, and style that influence how expressive or consistent a voice sounds. Choosing the right voice model and making small adjustments to these settings can significantly improve prosody without changing the script.
