Voice Formatting

What is Voice Input Formatted?

When interacting with VoiceCreator.ai assistants, you may notice references to "Voice Input Formatted" in call logs or system outputs. This feature enhances text-to-speech (TTS) interactions by cleaning and transforming raw text into a more natural format that is easier for the TTS provider to vocalise.

For example, it converts:

$57.75 → "fifty-seven dollars and seventy-five cents"
ST → "street"
Phone numbers → Spaced digits ("zero four one two three four five six seven eight nine")

Voice Input Formatted is enabled by default in your assistant’s voice provider settings to ensure high-quality voice outputs. However, you can disable this feature if you prefer raw, unmodified text outputs. We’ll guide you through this process below.

How Does It Work?

Voice Input Formatted performs a sequence of transformations to prepare text for TTS. It applies a series of helper functions that address specific text patterns. Here’s an overview of the step-by-step transformations:

Example Input

Let’s take the following raw text as an example:

Hi <tag> user,
**Please** call me at 0412-345-678,
price: $57.75
The date is 2025 03 15
time is 16:30
The address is 123 ST.
Email me at [email protected]

Step-by-Step Transformations

1. Remove Angle Bracket Content

Removes content within < > unless it contains <break>, <spell>, or double angle brackets << >>.
Example: <tag> is removed.

Result:

Hi  user,
**Please** call me at 0412-345-678,
price: $57.75
The date is 2025 03 15
time is 16:30
The address is 123 ST.
Email me at [email protected]

2. Remove Markdown Symbols

Strips characters like _, ~, and triple backticks, and optionally removes ** or * blocks.
Example: **Please** becomes Please.

Result:

Hi  user,
Please call me at 0412-345-678,
price: $57.75
The date is 2025 03 15
time is 16:30
The address is 123 ST.
Email me at [email protected]

3. Replace Newlines with Periods

Converts line breaks to periods and removes redundant periods.

Result:

Hi  user. Please call me at 0412-345-678. Price: $57.75. The date is 2025 03 15. Time is 16:30. The address is 123 ST. Email me at [email protected].

4. Replace Colons with Periods

Replaces : with .
Example: price: $57.75 becomes price. $57.75.

Result:

Hi  user. Please call me at 0412-345-678. Price. $57.75. The date is 2025 03 15. Time is 16:30. The address is 123 ST. Email me at [email protected].

5. Format Acronyms

Converts known acronyms to lowercase or spaces them out. Unrecognised acronyms with vowels remain unchanged.
Example: NASA → "nasa", but RD → "R D".

6. Format Dollar Amounts

Converts dollar amounts to words.
Example: $57.75 → "fifty-seven dollars and seventy-five cents".

7. Format Emails

Replaces @ with "at" and . with "dot".
Example: [email protected] → "jane dot doe at company dot com".

8. Format Dates

Converts YYYY MM DD to a spoken date format.
Example: 2025 03 15 → "Saturday, March 15, 2025".

9. Format Times

Converts 24-hour time to a readable format.
Example: 16:30 → "sixteen thirty".

10. Format Phone Numbers

Spaces out digits for clearer vocalisation.
Example: 0412-345-678 → "zero four one two three four five six seven eight nine".

11. Apply Replacements

Expands street suffixes and applies user-defined replacements.
Example: 123 ST → "123 street".

Customisation Options

Voice Input Formatted can be tailored to suit your needs:

1. Enabled

Enable or disable formatting globally.

Default: Enabled
To Disable:

"voice.chunkPlan.formatPlan.enabled": false

2. Number-to-Digits Cutoff

Defines when numbers are read as digits instead of words.

Default: Current year (2025).
Custom Example: To ensure numbers like 12345 are spelled out:

"numberToDigitsCutoff": 300000

3. Replacements

Configure specific word or pattern replacements using exact matches or regex rules.

Example:

Replace "hello" with "hi":

{ "type": "exact", "key": "hello", "value": "hi" }

Replace 5-letter words:

{ "type": "regex", "regex": "\\b[a-zA-Z]{5}\\b", "value": "hi" }

How to Disable Voice Formatting

To disable Voice Input Formatted, turn off the relevant settings:

"voice.chunkPlan.enabled": false

or:

"voice.chunkPlan.formatPlan.enabled": false

Conclusion

Voice Input Formatted transforms raw LLM outputs into natural and readable text for TTS. By fine-tuning or disabling these transformations, you can customise your assistant's behaviour to better meet your requirements.