> ## Documentation Index
> Fetch the complete documentation index at: https://developers.fastdrop.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Supported Languages

> Complete list of 99 supported transcription languages with accuracy tiers

FastDrop's transcription engine supports 99 languages. Accuracy varies significantly by language due to differences in training data availability. This page lists every supported language with its ISO code and expected accuracy tier.

## Accuracy tiers

Transcription accuracy depends on how much training data exists for each language. We classify languages into three tiers:

| Tier | Word Error Rate (WER) | What to expect |
| - | :-: | - |
| **High** | Under 8% | Production-ready. Accurate transcription with reliable word-level timestamps. Suitable for subtitles, search indexing, and automated workflows. |
| **Medium** | 8–15% | Good for most use cases. May occasionally miss uncommon words or proper nouns. Recommended to review output for professional use. |
| **Low** | 15–40%+ | Experimental. Useful for getting the gist of content, but expect frequent errors. Best paired with the `language` hint parameter for improved results. |

<Note>
  Word Error Rate (WER) measures the percentage of words transcribed incorrectly. A 5% WER means roughly 1 in 20 words may be wrong. These are approximate ranges — actual accuracy depends on audio quality, speaker clarity, background noise, and domain-specific vocabulary.
</Note>

## Tips for better accuracy

* **Provide a `language` hint** — Auto-detection works well for high-tier languages but can misidentify lower-resource languages. Passing the `language` parameter eliminates detection errors entirely.
* **Clean audio matters** — Background noise, music, and overlapping speakers reduce accuracy across all languages.
* **Medium and low-tier languages** benefit the most from a language hint. Without one, the detector may fall back to a related high-resource language.

## Complete language list

### High accuracy (under 8% WER)

These languages have extensive training data and deliver production-quality transcription.

| Language | ISO Code | Notes |
| - | :-: | - |
| English | `en` | Best overall accuracy |
| Chinese | `zh` | Mandarin Chinese; Cantonese may have reduced accuracy |
| German | `de` | |
| Spanish | `es` | Handles Latin American and European variants |
| French | `fr` | |
| Italian | `it` | |
| Japanese | `ja` | |
| Korean | `ko` | |
| Portuguese | `pt` | Brazilian and European |
| Dutch | `nl` | |
| Russian | `ru` | |
| Polish | `pl` | |
| Turkish | `tr` | |
| Swedish | `sv` | |
| Czech | `cs` | |
| Danish | `da` | |
| Finnish | `fi` | |
| Greek | `el` | |
| Romanian | `ro` | |
| Hungarian | `hu` | |
| Norwegian | `no` | Bokmål |
| Ukrainian | `uk` | |
| Indonesian | `id` | |
| Hindi | `hi` | |
| Vietnamese | `vi` | |
| Thai | `th` | |
| Hebrew | `he` | |
| Arabic | `ar` | Modern Standard Arabic; dialect accuracy varies |
| Catalan | `ca` | |
| Malay | `ms` | |

### Medium accuracy (8–15% WER)

These languages produce good results suitable for most workflows. A language hint is recommended.

| Language | ISO Code | Notes |
| - | :-: | - |
| Bulgarian | `bg` | |
| Croatian | `hr` | |
| Slovak | `sk` | |
| Slovenian | `sl` | |
| Lithuanian | `lt` | |
| Latvian | `lv` | |
| Estonian | `et` | |
| Serbian | `sr` | |
| Bosnian | `bs` | |
| Macedonian | `mk` | |
| Albanian | `sq` | |
| Persian (Farsi) | `fa` | |
| Urdu | `ur` | |
| Bengali | `bn` | |
| Tamil | `ta` | |
| Telugu | `te` | |
| Kannada | `kn` | |
| Malayalam | `ml` | |
| Marathi | `mr` | |
| Gujarati | `gu` | |
| Swahili | `sw` | |
| Tagalog | `tl` | |
| Azerbaijani | `az` | |
| Kazakh | `kk` | |
| Galician | `gl` | |
| Afrikaans | `af` | |
| Welsh | `cy` | |
| Icelandic | `is` | |
| Armenian | `hy` | |
| Georgian | `ka` | |
| Nepali | `ne` | |
| Punjabi | `pa` | |
| Sinhala | `si` | |
| Belarusian | `be` | |
| Nynorsk | `nn` | Norwegian Nynorsk |

### Low accuracy (15–40%+ WER)

These languages have limited training data. Transcription captures the general content but expect frequent errors. Always provide a `language` hint for best results.

| Language | ISO Code | Notes |
| - | :-: | - |
| Amharic | `am` | |
| Yoruba | `yo` | |
| Hausa | `ha` | Very limited training data |
| Somali | `so` | |
| Shona | `sn` | |
| Breton | `br` | |
| Basque | `eu` | |
| Occitan | `oc` | |
| Latin | `la` | |
| Mongolian | `mn` | |
| Khmer | `km` | |
| Lao | `lo` | |
| Burmese | `my` | |
| Tibetan | `bo` | |
| Uzbek | `uz` | |
| Turkmen | `tk` | |
| Tajik | `tg` | |
| Sindhi | `sd` | |
| Pashto | `ps` | |
| Maori | `mi` | |
| Sanskrit | `sa` | |
| Luxembourgish | `lb` | |
| Maltese | `mt` | |
| Yiddish | `yi` | |
| Haitian Creole | `ht` | |
| Faroese | `fo` | |
| Tatar | `tt` | |
| Assamese | `as` | |
| Malagasy | `mg` | |
| Lingala | `ln` | |
| Hawaiian | `haw` | Very limited training data |
| Bashkir | `ba` | |
| Javanese | `jw` | |
| Sundanese | `su` | |

## Language detection

When you omit the `language` parameter, FastDrop automatically detects the spoken language from the audio. Detection accuracy follows the same tier pattern:

* **High-tier languages** are detected reliably (95%+ accuracy)
* **Medium-tier languages** are detected correctly most of the time (85–95%)
* **Low-tier languages** may be misidentified as a related higher-resource language (60–85%)

For critical workflows, always pass the `language` parameter to bypass detection entirely.

## Translation coverage

Translation to English is available for all 99 languages. When `translate` is `true` (the default) and the source language is not English, FastDrop generates a full English translation alongside the source transcription.

Translation quality is generally high across all languages, as it operates on the transcribed text rather than raw audio. Even for low-accuracy transcription languages, the translation of whatever text *is* captured tends to be reliable.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.