Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.lyrcs.ai/llms.txt

Use this file to discover all available pages before exploring further.

lyrcs.ai supports 48 languages across two tiers. Indian languages have additional quality layers not present for global languages.

​
Indian languages — enhanced accuracy

All 30 Indian languages benefit from:
  • Language-specific prompt layers — additional context and instructions tuned for each language’s linguistic characteristics, script conventions, and common lyric patterns
  • Correction dictionaries — post-processing filters for common transcription errors specific to each language
  • Transliteration standards — romanisation follows consistent, widely-understood conventions (e.g. ITRANS-style for Hindi/Sanskrit, ISO 15919 influenced for Dravidian languages)
Indian languages include all 22 of India’s officially scheduled languages plus additional widely-spoken regional languages.

​
Global languages

The 18 global languages use the base transcription pipeline without language-specific correction layers. Quality is high for well-resourced languages (English, Spanish, French, Mandarin Chinese) and good for others.

​
Full language table

​
Indian Languages — 30

LanguageFamilyScript
HindiIndo-AryanDevanagari
BengaliIndo-AryanBengali
MarathiIndo-AryanDevanagari
TeluguDravidianTelugu
TamilDravidianTamil
GujaratiIndo-AryanGujarati
UrduIndo-AryanNastaliq (Perso-Arabic)
KannadaDravidianKannada
OdiaIndo-AryanOdia
MalayalamDravidianMalayalam
PunjabiIndo-AryanGurmukhi
AssameseIndo-AryanAssamese (Bengali variant)
MaithiliIndo-AryanDevanagari / Tirhuta
SantaliAustroasiaticOl Chiki
KashmiriIndo-AryanPerso-Arabic / Devanagari
NepaliIndo-AryanDevanagari
SindhiIndo-AryanPerso-Arabic / Devanagari
KonkaniIndo-AryanDevanagari / Latin
DogriIndo-AryanDevanagari
ManipuriSino-TibetanMeitei Mayek / Bengali
BodoSino-TibetanDevanagari
SanskritIndo-AryanDevanagari
HaryanviIndo-AryanDevanagari
BhojpuriIndo-AryanDevanagari
MagahiIndo-AryanDevanagari
ChhattisgarhiIndo-AryanDevanagari
RajasthaniIndo-AryanDevanagari
TuluDravidianKannada / Tigalari
GondiDravidianDevanagari / Gondi
SaurashtraIndo-AryanSaurashtra / Latin

​
Global Languages — 18

LanguageFamily
EnglishGermanic
SpanishRomance
Mandarin ChineseSino-Tibetan
FrenchRomance
ArabicSemitic
PortugueseRomance
RussianSlavic
JapaneseJaponic
GermanGermanic
KoreanKoreanic
ItalianRomance
TurkishTurkic
DutchGermanic
IndonesianAustronesian
ThaiTai-Kadai
YorubaNiger-Congo
SwahiliNiger-Congo
ZuluNiger-Congo
The script field is not returned by GET /api/v1/languages — only name and family. The script column above is for reference only. The transcript output uses whatever script the audio contains, as recognised by the model.

​
Using language values

Always use the exact name strings from GET /api/v1/languages. Values are case-sensitive. 
# Correct
"language": "Mandarin Chinese"

# Wrong — will return 400
"language": "mandarin chinese"
"language": "Chinese"
"language": "zh"

​
Known limitations

  • WAV/FLAC duration reporting: Audio submitted as WAV or FLAC may show duration_seconds: 0 in job responses. This is a metadata extraction issue and does not affect transcript or alignment quality.
  • Script variation: For languages with multiple scripts (e.g. Kashmiri in Perso-Arabic vs Devanagari), the output script is determined by what the model detects in the audio and context. If you need a specific script, contact api@lyrcs.ai.