# Last Mile Road

> Speech-to-text for underserved languages.

Last Mile Road provides affordable, pay-as-you-go speech-to-text and translation for over 700 languages that commercial platforms overlook. We serve NGOs, researchers, programme evaluators, and human language professionals working in low-resource languages across the Pacific, Africa, South Asia, and Southeast Asia.

## Key facts

- 700+ languages covered and growing
- $0.15/minute for transcription and translation combined
- 20-minute free trial, no credit card required
- Pay-as-you-go, no subscriptions, no monthly minimum
- Audio files processed on own infrastructure, never sent to third-party APIs
- Customer data is never used for model training unless explicitly opted in
- Up to 150 minutes per audio file
- Supports MP3, WAV, M4A, AAC formats
- Output formats: DOCX, Markdown, plain text
- Speaker diarization and timestamps included
- Translation to 20 target languages for Pacific languages, English for Indic
- Built on Meta Omnilingual ASR (Apache 2.0) + Gemma 4 / Krutrim translation

## Languages with Good accuracy (CER < 1%)

- Tok Pisin (tpi) — Papua New Guinea
- Solomon Pijin (pis) — Solomon Islands
- Fijian (fij) — Fiji
- Kikuyu (kik) — Kenya
- Sango (sag) — Central Africa
- Ilocano (ilo) — Philippines
- Bemba (bem) — Zambia/DRC
- Faroese (fao) — Faroe Islands
- Jamaican Patois (jam) — Caribbean
- Papiamento (pap) — Aruba/Caribbean
- Hiligaynon (hil) — Philippines
- Sundanese (sun) — Indonesia
- And 600+ more in the Good tier

## Languages with Fair accuracy (CER > 1%, light editing needed)

- Swahili (swh) — East Africa
- Hausa (hau) — West Africa
- Bengali (ben) — Bangladesh/India
- Hindi (hin) — India
- Amharic (amh) — Ethiopia
- Nepali (nep) — Nepal
- Tamil (tam) — India
- Vietnamese (vie) — Southeast Asia

## Use cases

- Health research field interviews in local languages
- Most Significant Change (MSC) story collection for programme evaluation
- Participatory community research and focus group transcription
- Human translator draft generation and workflow acceleration

## Trust and privacy

- Audio processed on own servers and cloud GPU infrastructure (Modal)
- No training on customer data. Period.
- Automatic audio deletion after processing
- Built on Meta Omnilingual ASR (Apache 2.0)
- See full privacy policy, terms, and refund policy on the site

## Pages

- Home: https://transcribe.lastmileroad.org/
- How it works: https://transcribe.lastmileroad.org/#how-it-works
- Languages: https://transcribe.lastmileroad.org/#languages
- Pricing: https://transcribe.lastmileroad.org/#pricing
- About: https://transcribe.lastmileroad.org/#mission
- Login: https://transcribe.lastmileroad.org/login
- Register: https://transcribe.lastmileroad.org/register
- Privacy Policy: https://transcribe.lastmileroad.org/privacy.html
- Terms of Service: https://transcribe.lastmileroad.org/terms.html
- Refund Policy: https://transcribe.lastmileroad.org/refund.html