Introducing FastVision AI: OCR and TTS you actually own

Every OCR vendor pitch starts the same way: upload your documents to our cloud. For a huge class of teams — hospitals, banks, law firms, governments — that sentence ends the conversation. The documents are the sensitive part. They cannot leave.

FastVision AI is our answer: a complete OCR and Text-to-Speech platform that installs with docker compose and runs entirely inside your network. It looks and feels like a modern API SaaS — dashboard, API keys, webhooks, usage metering, SDKs in six languages — but the models run on your machines.

What's in the box

• Document OCR built on PaddleOCR with DocTR as an automatic fallback, plus OpenCV preprocessing for hostile scans
• Purpose-tuned extraction for receipts, invoices, passports, ID cards, bank statements and business cards
• Bounding boxes and per-block confidence on every result, plus searchable-PDF output
• Neural Text-to-Speech in 9 languages powered by Piper, with chunked streaming and SSML
• Async job architecture: 202 on submit, Celery workers, signed webhooks on completion
• SDKs for JavaScript, Python, Go, Swift, Kotlin and Flutter

Why self-hosted, why now

Three forces converged. Open OCR models crossed the quality threshold where they beat cloud APIs on real-world documents. CPU inference got fast enough that a modest box handles serious volume. And regulators sharpened their teeth — GDPR, HIPAA and a wave of data-residency laws have made 'just send it to a US cloud API' a legal project, not an HTTP call.

The best data-protection agreement is the one you never have to sign, because the data never left.

Five minutes to your first extraction

git clone https://github.com/fastvision-ai/fastvision
cd fastvision
docker compose up -d
# dashboard on :3000, API on :8000

curl -X POST http://localhost:8000/api/v1/ocr/image \
  -H "X-API-Key: fv_live_yourkeyhere" \
  -F "file=@receipt.png" -F "doc_type=receipt"

The free plan includes 100 OCR pages and 10,000 TTS characters every month — enough to evaluate seriously. Paid plans start at $19. Enterprise teams get air-gapped installs, SSO and an SLA with a signature on it.

We built FastVision for the teams everyone else's terms of service quietly exclude. If that's you, we'd love to see what you build.