# Gladia

Speech-to-text API for pre-recorded and live audio that returns transcripts with timestamps, speaker labels and optional add-ons.

- Provider: Gladia (https://www.gladia.io)
- Category: Transcribe audio
- Type: Hosted API
- Pricing model: Prepaid credit wallet, billed per hour of audio [1]
- Free tier: EUR 50 in credits, one time, valid for one year [1]
- Authentication: API key in the `x-gladia-key` header [3]
- Connect via: REST API, WebSocket, JavaScript and Python SDKs, CLI, local MCP server [16]
- Read from the provider's pages on 2026-11-04
- Page: https://supermarket.dev/tools/transcribe-audio/gladia

## Pricing

### Billing
- Pricing model: Prepaid credits with no subscription; pay for the audio processed [1]
- Free tier: EUR 50 in credits, one time, valid for one year; free accounts get 3 concurrent pre-recorded jobs and 1 live session [1]
- Card required for free tier: Not published
- Minimum spend: Not published
- Billing period: No billing cycle on Self-serve: credits are topped up manually or by auto top-up. Enterprise is an annual plan [1]

### Prices
Prices are in euros (EUR), not converted. Per-hour rates are as published. The default model for a request is `solaria-1`, so the first row is the default and the third is the advertised "starts at" price.
- Async (pre-recorded), `solaria-1` (default model): EUR 0.27 per hour of audio at 0-100 hours a month, falling to EUR 0.21 per hour above 10,000 hours (Self-serve) [1]
- Real-time (live), `solaria-1`: EUR 0.68 per hour of audio (Self-serve) [1]
- Async (pre-recorded), `solaria-3` (advertised price): EUR 0.16 per hour of audio at 0-100 hours a month, falling to EUR 0.12 per hour above 10,000 hours. Used only if `model` is set to `solaria-3`; pre-recorded, 5 languages [1]
- Advertised starting price: From EUR 0.16 per hour of async audio on `solaria-3` [1]

### Async rate by monthly volume
| Monthly volume | `solaria-1` EUR per hour (default) | `solaria-3` EUR per hour |
|---|---|---|
| 0 - 100 hours | 0.27 | 0.16 |
| 100 - 500 hours | 0.26 | 0.15 |
| 500 - 2,000 hours | 0.24 | 0.14 |
| 2,000 - 5,000 hours | 0.23 | 0.13 |
| 5,000 - 10,000 hours | 0.22 | 0.13 |
| 10,000+ hours | 0.21 | 0.12 |
Source [1]

### Other charges
Core features such as custom vocabulary are included in the base rate. Top-ups of EUR 1,000 or more earn bonus credits (+10% at EUR 1,000, +15% at EUR 2,500, +20% at EUR 5,000).
| Add-on | Async, EUR per hour | Real-time, EUR per hour |
|---|---|---|
| Summarization | 0.03 | 0.03 |
| Sentiment analysis | 0.02 | 0.02 |
| Translation | 0.05 | 0.05 |
| Named entity recognition | 0.07 | 0.07 |
| PII redaction | 0.07 | Not available |
| Audio to LLM | From EUR 0.20 in / 1.25 out per 1M tokens | Not available |
Source [1]

### Plans
| Plan | Price | Concurrency | Notes |
|---|---|---|---|
| Self-serve | Pay as you go, see Prices | 25 async, 30 real-time | Model training opt-in by default, free opt-out |
| Enterprise | Custom | Unlimited | Custom models, zero data retention, custom hosting, SLAs |
Source [1]

## Tools

### MCP server
- Official MCP server: Yes, local [12]
- Address or package: `@gladiaio/mcp` on npm, run with `npx -y @gladiaio/mcp`; source in the `gladiaio/gladia-mcp` GitHub repository [12]

### MCP tools
Read-only column comes from the `readOnlyHint` annotations in the server source (`src/server.ts`). The server does not start or stream live audio.
| Tool | What it does | Read-only |
|---|---|---|
| `transcribe_pre_recorded` | Starts an asynchronous, billable pre-recorded job | No |
| `get_pre_recorded_transcription` | Gets a pre-recorded job's current status and available result | Yes |
| `list_pre_recorded_transcriptions` | Lists pre-recorded jobs with optional filters and pagination | Yes |
| `delete_pre_recorded_transcription` | Permanently deletes a pre-recorded job and its data | No |
| `upload_audio` | Uploads a file from an allowed root (Desktop by default) | No |
| `get_live_transcription` | Gets an existing live session's status and available result | Yes |
| `list_live_transcriptions` | Lists existing live sessions with optional filters and pagination | Yes |
| `delete_live_transcription` | Permanently deletes an existing live session and its data | No |
Source [12]

### API
Endpoint list taken from the OpenAPI spec; `/v2/transcription` routes are deprecated.
| Endpoint | What it does |
|---|---|
| `POST /v2/upload` | Upload an audio file for a pre-recorded job |
| `POST /v2/pre-recorded` | Start a pre-recorded transcription job |
| `GET /v2/pre-recorded/{id}` | Get status, parameters and result of a pre-recorded job |
| `GET /v2/pre-recorded` | List pre-recorded transcriptions |
| `DELETE /v2/pre-recorded/{id}` | Delete a pre-recorded transcription and its data |
| `POST /v2/live` | Start a live session; the response is used to open a WebSocket |
| `GET /v2/live/{id}` | Get status, parameters and result of a live transcription |
| `GET /v2/live/{id}/file` | Download the audio recorded during a live transcription |
Source [21]

### Features
- Default model: `solaria-1` [5]
- Default model (other model): `solaria-3` must be set with `model`: pre-recorded only, no code switching, English, French, German, Spanish and Italian [2]
- Languages: 100+ languages with `solaria-1` (default), including code switching [5]
- Languages (`solaria-3`): 5: English, French, German, Spanish, Italian [2]
- Speaker labels: Pre-recorded only, via `diarization`; speakers get an index in order of first appearance [10]
- Word timestamps: Yes, on by default in pre-recorded results [11]
- Live streaming: Yes, over WebSocket, `solaria-1` only; one session can last up to 3 hours [2]
- Max file size: 1000 MB and 135 minutes per pre-recorded request (4h15 on Enterprise) [7]
- Audio input: Pre-recorded: URL or uploaded file; audio (aac, ac3, eac3, flac, m4a, mp2, mp3, ogg, opus, wav) and video (mp4, mov, mkv and others); live: raw `wav/pcm`, `wav/alaw` or `wav/ulaw` chunks [7]
- Audio input (live encodings): `wav/pcm`, `wav/alaw`, `wav/ulaw` [18]

## Authentication

### Methods
| Method | Used for | How |
|---|---|---|
| API key | All REST and WebSocket-init endpoints | Send the key in the `x-gladia-key` header |
| API key in environment | SDKs, CLI and the local MCP server | Set `GLADIA_API_KEY`; the CLI also has `gladia auth set <key>` |
Source [3]

### Getting access
- Who creates the account: A person signs up at app.gladia.io (Google sign-in); sign-up by an agent alone is not described [3]
- Where to get a key: API keys section of app.gladia.io; a default key is created at sign-up [3]
- Other requirements: Not published

## Setup

### Addresses
- REST API: `https://api.gladia.io/v2/pre-recorded` [2]
- MCP server: No hosted address; local stdio server run with `npx -y @gladiaio/mcp` [12]
- OpenAPI spec: `https://api.gladia.io/openapi.json` [22]
- Docs for agents: `https://docs.gladia.io/llms.txt`; every docs page also has a `.md` copy [21]
- Status page: `https://status.gladia.io/` [19]

### Install
Rows come from the SDK, MCP README, CLI and Gladia Skills pages (sources `sdk`, `mcp`, `cli`, `skills`).
| What | Command |
|---|---|
| MCP server, Claude Code | `claude mcp add gladia --env GLADIA_API_KEY=your-api-key -- npx -y @gladiaio/mcp` |
| MCP server, Codex | `codex mcp add gladia --env GLADIA_API_KEY=your-api-key -- npx -y @gladiaio/mcp` |
| MCP server, Cursor and Claude Desktop | JSON config with `"command": "npx"` and `"args": ["-y", "@gladiaio/mcp"]` |
| JavaScript SDK | `npm install @gladiaio/sdk` |
| Python SDK | `pip install gladiaio-sdk` |
| CLI, macOS and Linux | `curl -fsSL https://github.com/gladiaio/gladia-cli/releases/latest/download/install.sh \| sh` |
| Agent skills, Claude Code | `/plugin marketplace add gladiaio/skills` then `/plugin install gladia` |
| Agent skills, any agent | `npx skills add gladiaio/skills` |
Source [16]

## Limits

### Rate limits
Exceeding concurrency returns HTTP 429. A separate per-user API request rate limit exists but its number is not published.
| Plan | Concurrent pre-recorded jobs | Concurrent live sessions | Queued async requests |
|---|---|---|---|
| Free | 3 | 1 | Not stated |
| Paid | 25 | 30 | Up to 300 |
| Enterprise | On demand | On demand | Not stated |
Source [6]

### Other limits
- Pre-recorded audio length: 135 minutes per request; 4h15 on Enterprise [7]
- Pre-recorded file size: 1000 MB [7]
- Live session length: 3 hours per session [6]
- Channels: Pre-recorded: up to 2; live: up to 8 [7]
- API request rate (calls per time frame): Not published

## Privacy

### Your data
- Data location: EU and US workloads are kept separate on region-specific infrastructure; HDS hosting for enterprise [20]
- Retention: Paid accounts: audio and transcripts 3 weeks, request metadata 1 year. Free accounts: all data 1 year. Enterprise can set a custom period [8]
- Used for training: Self-serve: opted in by default, free opt-out from the profile; Enterprise: opted out by default [1]
- Used for training (other page): Never used for model training unless explicitly agreed [20]
- Zero data retention: Enterprise only; file upload and result polling are disabled, results come by callback [8]
- Zero data retention (other page): Described as a policy for paid plans [20]

### Compliance
- Certifications: SOC 2 Type 2, ISO 27001, GDPR [1]
- HIPAA: Listed as HIPAA compliant; Business Associate Agreements available [20]

## Sources

1. Pricing: https://www.gladia.io/pricing
2. Models: https://docs.gladia.io/chapters/introduction/models.md
3. Authentication: https://docs.gladia.io/api-reference/authentication.md
4. Getting started: https://docs.gladia.io/chapters/introduction/getting-started.md
5. Pre-recorded quickstart: https://docs.gladia.io/chapters/pre-recorded-stt/quickstart.md
6. API rate limits and transcription concurrency: https://docs.gladia.io/chapters/limits-and-specifications/concurrency.md
7. Supported files and duration: https://docs.gladia.io/chapters/limits-and-specifications/supported-formats.md
8. Data retention and zero data retention: https://docs.gladia.io/chapters/limits-and-specifications/data-retention.md
9. Supported languages: https://docs.gladia.io/chapters/language/supported-languages.md
10. Speaker diarization: https://docs.gladia.io/chapters/audio-intelligence/speaker-diarization.md
11. Word-level timestamps: https://docs.gladia.io/chapters/pre-recorded-stt/features/word-level-timestamps.md
12. Gladia MCP README: https://raw.githubusercontent.com/gladiaio/gladia-mcp/main/README.md
13. Gladia MCP server source: https://raw.githubusercontent.com/gladiaio/gladia-mcp/main/src/server.ts
14. Gladia Skills README: https://raw.githubusercontent.com/gladiaio/skills/main/README.md
15. CLI: https://docs.gladia.io/chapters/how-to-use-gladia/cli.md
16. SDK: https://docs.gladia.io/chapters/how-to-use-gladia/sdk.md
17. Streaming quickstart: https://docs.gladia.io/chapters/live-stt/quickstart.md
18. OpenAPI spec: https://api.gladia.io/openapi.json
19. Status page: https://status.gladia.io/
20. Compliance hub: https://www.gladia.io/compliance-hub
21. llms.txt: https://docs.gladia.io/llms.txt
22. Checked by supermarket.dev