# Privocio — Private Speech-to-Text API

> Privacy-first, flat-rate speech-to-text API for AI agents, developers, and teams that need accurate transcription without per-minute billing or data exposure.

Privocio is a hosted and self-hosted speech-to-text API designed as a drop-in replacement for the OpenAI Whisper API. It offers fixed-price plans instead of per-minute billing, three output modes optimized for different workflows, and a strict privacy-first architecture where audio and transcripts are never used for model training.

## Why Privocio

- **Privacy-first**: Audio and transcripts are never shared, sold, or used to train models. Self-hosted deployment available for full data sovereignty.
- **Predictable pricing**: Flat-rate plans billed every 4 weeks — no per-minute charges, no overage fees, no surprise bills.
- **OpenAI SDK compatible**: Works as a drop-in replacement for the OpenAI Whisper API. Change `base_url` and you're done.
- **Token optimization**: Agent output mode reduces downstream LLM token usage by approximately 60% by removing filler words, silence markers, and transcript noise before the text reaches your model.
- **Self-hosted option**: Deploy on your own infrastructure for teams with strict compliance, data residency, or air-gapped requirements.

## Plans and Pricing

| Plan | Price | Audio Allowance | API Access | Key Limits |
|------|-------|-----------------|------------|------------|
| Free | $0 | 3 hours / 4 weeks | Browser only | 3 RPM, 1 concurrent, 3 min/request, 10 MB uploads |
| Go | $19 / 4 weeks | 400 hours / 4 weeks | API key | 30 RPM, 2 concurrent, 15 min/request, 100 MB uploads |
| Pro | $39 / 4 weeks | 800 hours / 4 weeks | API key | 30 RPM, 2 concurrent, 15 min/request, 100 MB uploads, team members |
| Enterprise | Custom | Unlimited | API key + SSO | Custom limits, self-hosted, dedicated support, SLAs |

Quarterly billing available: Go $51.30 / 3 months, Pro $105.30 / 3 months.

## Output Modes

- **Raw**: Full verbatim transcript with all filler words, false starts, and repetitions preserved. Best for legal, compliance, and archival use cases.
- **Clean**: Noise removed, punctuation corrected, readability improved. Best for meeting notes, content creation, and human consumption.
- **Agent**: Structured JSON output with token-efficient formatting optimized for AI agent consumption. Removes transcript noise before it reaches your LLM, reducing context size and downstream token costs by ~60%.

## Technical Specifications

- API base URL: `https://api.privocio.com`
- Authentication: Bearer token (API key or session token)
- Supported audio formats: MP3, WAV, M4A, FLAC, OGG, WebM
- Supported languages: 100 with automatic language detection
- Max audio per request: 15 minutes (paid plans), 3 minutes (free)
- Max upload size: 100 MB (paid plans), 10 MB (free)
- Response format: JSON (full text, segments with start/end timestamps, speaker labels)
- Streaming: Server-Sent Events (SSE) with real-time segment delivery
- Speaker diarization: Automatic speaker separation with configurable speaker count
- SDK compatibility: Works with OpenAI Python SDK (`openai.OpenAI(base_url="https://api.privocio.com/v1")`) and OpenAI Node.js SDK
- Latency: ~420ms p95 end-to-end for typical audio segments
- Hosted uptime: 99.9% uptime SLA target on Go and Pro plans; live status at https://privocio.com/status
- Health endpoints: GET https://api.privocio.com/healthz (liveness), GET https://api.privocio.com/v1/status (component health)
- Webhooks: Async callback support for batch processing workflows

## Integration Example

```python
from openai import OpenAI

client = OpenAI(
    api_key="your-privocio-api-key",
    base_url="https://api.privocio.com/v1"
)

transcription = client.audio.transcriptions.create(
    model="whisper-1",
    file=open("meeting.mp3", "rb"),
    response_format="json"
)
```

## Who Should Use Privocio

- **AI agent builders**: Need structured, clean voice input for agent frameworks (LangChain, CrewAI, AutoGen). Agent mode produces token-efficient JSON that reduces LLM costs.
- **Developers building voice features**: Need a simple, well-documented API with predictable costs. OpenAI SDK compatibility means minimal integration work.
- **Privacy-sensitive teams**: Healthcare, legal, finance, and government teams that cannot send audio to shared cloud APIs. Self-hosted deployment ensures data never leaves your infrastructure.
- **High-volume transcription users**: Teams processing 50+ hours/month where per-minute pricing becomes expensive. Flat-rate plans cap costs regardless of usage within the allowance.
- **Product teams**: Adding voice workflows to SaaS products without exposing users to variable usage costs or third-party data processors.

## Comparison with Alternatives

| | Privocio | OpenAI Whisper API | Google STT | AWS Transcribe |
|---|---|---|---|---|
| Pricing model | Flat-rate | Per-minute ($0.006/min) | Per-minute | Per-second |
| Data used for training | Never | Possible | Possible | Possible |
| Self-hosted option | Yes | No | No | No |
| Token optimization | Built-in | None | None | None |
| OpenAI SDK compatible | Yes (drop-in) | Native | No | No |
| Streaming | SSE | No | Yes | Yes |
| Speaker diarization | Yes | No | Yes | Yes |

At 200 hours/month, OpenAI Whisper API costs ~$72/month. Privocio Pro covers 800 hours for $39/4 weeks.

## Use Cases

- **AI Agents**: Voice commands → structured JSON → agent framework. Reduces token costs and eliminates transcript noise from agent context windows.
- **Healthcare**: HIPAA-aligned transcription for clinical notes, telemedicine, medical dictation, and patient intake. Self-hosted deployment for full compliance.
- **Finance**: Secure voice processing for trading floors, client calls, and compliance workflows. Self-hosted options and retention controls for regulated data.
- **Legal**: Privileged audio stays in your perimeter. Transcription for depositions, client intake, and matter workflows without training on your data.
- **Developers**: Build voice features with a familiar API. Drop-in OpenAI SDK replacement with better pricing and privacy.
- **Meeting transcription**: Real-time streaming with speaker diarization for meeting notes, action items, and searchable archives.
- **Content creation**: Clean mode produces publication-ready transcripts from interviews, podcasts, and video narration.
- **Customer support**: Transcribe support calls for quality assurance, training, and compliance with automatic speaker separation.

## Core

- [Homepage](https://privocio.com): Private, flat-rate speech-to-text API overview.
- [Pricing](https://privocio.com/pricing): Flat-rate plans — Free, Go ($19/4 weeks), Pro ($39/4 weeks), Enterprise.
- [API Documentation](https://privocio.com/docs): Authentication, transcription endpoints, streaming, rate limits.
- [Free Transcription Tool](https://privocio.com/free-speech-to-text-stt-online-tool): In-browser audio-to-text, no signup required.
- [Supported Languages](https://privocio.com/en/languages): 100+ languages with automatic detection.
- [Blog](https://privocio.com/blog): Guides and technical notes on STT, AI agents, and token optimization.
- [Security](https://privocio.com/security): Encryption, retention controls, and no model training on customer audio.
- [System Status](https://privocio.com/status): Live API uptime and component health.

## Use cases

- [Use Cases Hub](https://privocio.com/use-cases): All speech-to-text use cases.
- [AI Agents](https://privocio.com/use-cases/ai-agents): Structured voice input and token-optimized output for agent frameworks.
- [Healthcare](https://privocio.com/use-cases/healthcare): Private, self-hostable transcription for clinical workflows.
- [Developers](https://privocio.com/use-cases/developers): OpenAI-compatible STT with predictable pricing.
- [Finance](https://privocio.com/use-cases/finance): Secure voice processing with retention controls.
- [Legal](https://privocio.com/use-cases/legal): Privileged audio transcription that stays in your perimeter.

## SDKs and integrations

- [SDK Hub](https://privocio.com/sdk): Python and JavaScript speech-to-text SDKs.
- [Python SDK](https://privocio.com/sdk/python): OpenAI-compatible Python transcription guide.
- [JavaScript SDK](https://privocio.com/sdk/javascript): Node and browser transcription guide.
- [Integrations Hub](https://privocio.com/integrations): Integration patterns for privacy-first voice workflows.
- [OpenAI SDK Integration](https://privocio.com/integrations/openai-sdk): Drop-in OpenAI Whisper replacement via baseURL.
- [LangChain Integration](https://privocio.com/integrations/langchain): Wire Privocio STT into LangChain agents.

## Comparisons and guides

- [Comparison Hub](https://privocio.com/compare): Compare Privocio to other speech-to-text APIs.
- [Privocio vs OpenAI Whisper](https://privocio.com/compare/privocio-vs-openai-whisper): Pricing, privacy, and SDK compatibility.
- [Privocio vs AWS Transcribe](https://privocio.com/compare/privocio-vs-aws-transcribe): Flat-rate vs per-second billing.
- [Privocio vs Google STT](https://privocio.com/compare/privocio-vs-google-speech-to-text): TCO and privacy-first deployment.
- [Migrate from OpenAI Whisper](https://privocio.com/guides/migrate-from-openai-whisper): One-line migration guide.

## Blog guides

- [Python STT tutorial](https://privocio.com/blog/python-speech-to-text-api): Transcribe audio in Python.
- [JavaScript STT tutorial](https://privocio.com/blog/javascript-speech-to-text-api): Transcribe audio in JavaScript.
- [OpenAI Whisper alternative](https://privocio.com/blog/openai-whisper-api-alternative): Why and how to switch.
- [LangChain voice agent](https://privocio.com/blog/langchain-voice-agent-privocio): Build a voice agent with LangChain.

## Legal and contact

- [Privacy Policy](https://privocio.com/legal/privacy-policy): How Privocio handles personal data.
- [Terms of Service](https://privocio.com/legal/terms-of-service): Terms for using Privocio services.
- [Contact](mailto:hello@privocio.com): Email the Privocio team.

## Company

Privocio is operated by The Useless Company PTE LTD. The product is designed and built for teams and developers who believe voice infrastructure should be private, predictable, and developer-friendly.