Embedding Custom Phonemes in Async TTS API

1 Why provide custom phonemes?

Default grapheme‑to‑phoneme (G2P) conversion works well for most English words, but brand names, acronyms, foreign words, and creative spellings can trip it up. The <phonemes> tag lets you override the pronunciation inline, so you keep the surrounding text natural while guaranteeing a correct reading of tricky words.

2 Tag syntax

Element	Required	Description
`<phonemes>`	✓	Opening tag marking the start of the phonetic override.
`…IPA string…`	✓	A sequence of IPA symbols. Separate words with spaces. Use `ˈ` (primary stress) and `ˌ` (secondary stress) as needed.
`</phonemes>`	✓	Closing tag.

Example
Welcome to <phonemes>ˈeɪ.sɪŋk</phonemes>.

3 Supported symbol set

Standard IPA characters for English
(e.g., ɹ, ʒ, ð, ŋ, etc.).

Stress marks: ˈ (primary) & ˌ (secondary).

Word separator: space ( ). Do not use slashes (/…/) or brackets ([…]).

Unsupported at present:

Tone letters, diacritics outside the English subset.

SAMPA, ARPABET or other non‑IPA alphabets.

4 Request format

POST /text_to_speech/streaming HTTP/1.1
Host: api.async.ai
Content-Type: application/json
X-Api-Key: <YOUR_API_KEY>

{
  "model_id": "asyncflow_v2.0",
  "transcript": "Welcome to <phonemes>ˈeɪ.sɪŋk</phonemes>. Here you can test our text-to-speech capabilities.",
  "voice": { "mode": "id", "id": "e0f39dc4-f691-4e78-bba5-5c636692cc04" },
  "output_format": {
    "container": "raw",
    "encoding": "pcm_s16le",
    "sample_rate": 44100
  }
}

Curl shortcut

5 Best practices

Keep tags short – one or two words. Large phoneme blocks reduce prosody quality.

Match language – only supply English IPA for now; mixed‑language input may degrade output.

Stress wisely – adding primary stress (ˈ) usually suffices.

Test incrementally – start with small examples and listen before batch‑generating.

6 English IPA Symbol Reference

Below is a quick‑lookup table of the standard IPA symbols you can embed inside <phonemes> when working with English voices in Async. All examples use General American pronunciations.

6.1 Consonants

IPA	Sample word	Full IPA
p	pin	pɪn
b	bin	bɪn
t	tin	tɪn
d	din	dɪn
k	cat	kæt
ɡ	go	ɡoʊ
m	man	mæn
n	no	noʊ
ŋ	sing	sɪŋ
f	fan	fæn
v	van	væn
θ	thin	θɪn
ð	this	ðɪs
s	see	siː
z	zoo	zuː
ʃ	she	ʃiː
ʒ	measure	ˈmɛʒɚ
h	he	hiː
tʃ	check	tʃɛk
dʒ	jump	dʒʌmp
r	red	ɹɛd
l	let	lɛt
j	yes	jɛs
w	we	wiː

6.2 Vowels

6.2.1 Monophthongs

IPA	Sample word	Full IPA
i	see	siː
ɪ	sit	sɪt
e	say (Canadian/British)	seɪ
ɛ	set	sɛt
æ	cat	kæt
ʌ	cut	kʌt
ə	about	əˈbaʊt
ɜː	bird (British)	bɜːd
ɑ	father	ˈfɑðɚ
ɔ	talk	tɔk
ʊ	foot	fʊt
u	food	fuːd

6.2.2 Diphthongs

IPA	Sample word	Full IPA
eɪ	say	seɪ
aɪ	sigh	saɪ
ɔɪ	joy	dʒɔɪ
aʊ	sow	saʊ
oʊ	so	soʊ
ɪər	here (British)	hɪər
eər	bear (British)	beər
ʊər	tour (British)	tʊər

6.3 Suprasegmentals & Diacritics

IPA	Meaning	Example	Notes
ˈ	Primary stress	banana → bəˈnænə	Mark the vowel beginning the stressed syllable.
ˌ	Secondary stress	transporˌtaˈtion	Optional.
ː	Length mark	see → siː	Indicates vowel length; appears in many vowels above.

7 FAQ

Q Can I use SAMPA or ARPABET?
A Not yet. We currently accept only IPA symbols for English.

Q Do I have to tag the whole sentence?
A No. Tag only the parts you want to override; everything else is phonemized automatically.

Q Will you support other languages?
A Yes—multilingual phoneme overrides are on the roadmap. Stay tuned!

Embedding Custom Phonemes in Async TTS API

1 Why provide custom phonemes?#

2 Tag syntax#

3 Supported symbol set#

4 Request format#

Curl shortcut#

5 Best practices#

6 English IPA Symbol Reference#

6.1 Consonants#

6.2 Vowels#

6.2.1 Monophthongs#

6.2.2 Diphthongs#

6.3 Suprasegmentals & Diacritics#

7 FAQ#