[Integrations](https://spiral.ai/integrations) › [Windsurf](https://spiral.ai/integrations/windsurf) › [Nova-3](https://spiral.ai/integrations/windsurf/nova-3)

# How to Use Nova-3 in Windsurf

Nova-3 transcribes meetings, interviews, and noisy recordings into text. It combines multilingual transcription with vocabulary hints for product names and technical terms.

|  |  |
| --- | --- |
| Model | Nova-3 |
| Available through | Deepgram |
| Category | Audio |
| Price | $0.00559 per minute of audio, per channel ($0.00676 with language multi). Redaction: $0.0026 per minute. Key terms: $0.00169 per minute. Entity detection: $0.00221 per minute. |
| Returns | json |
| Runs | In the background |
| Setup | About 2 minutes, no API key |

## Prompts to try with Nova-3 in Windsurf

- Use Nova-3 to transcribe the recorded product demo linked in this project's README. Choose up to 30 product and technical terms from the repository as vocabulary hints, keep the list below 500 tokens, and draft a transcript with timestamps.
- Use Nova-3 to transcribe the tutorial audio linked in docs/tutorials.md. Convert the timed words into draft SRT captions and save them beside the tutorial.
- Use Nova-3 to transcribe the architecture discussion linked in this issue. Separate speakers, then draft a decision record with a timestamp for each supporting quote.
- Use Nova-3 to transcribe the multilingual demo linked in the release notes. Preserve the spoken languages and create a timestamped transcript for review.
- Use Nova-3 to transcribe the customer interview linked in research/interviews.md. Separate speakers and draft a list of product requests, each backed by a quote and timestamp.

## How to set up Nova-3 in Windsurf

Pick the way that suits you. Each takes about two minutes.

### Use Spiral's MCP server

1. **Add the server**: Open the Customize page from the Agents window and click MCPs. Under This computer, open the config file, or edit it directly: `~/.config/devin/mcp_config.json` on macOS and Linux, `%APPDATA%\devin\mcp_config.json` on Windows. Add this:

```json
{
  "mcpServers": {
    "spiral": {
      "url": "https://api.spiral.ai/mcp"
    }
  }
}
```

2. **Sign in**: Click Authenticate on the Spiral server in Customize › MCPs, or let the agent call Spiral once: your browser opens. Sign in to Spiral, pick the organization and click Authorize. No API key.

```bash
devin mcp login spiral
```

3. **Check it**: Spiral is listed under This computer in Customize › MCPs. From the terminal:

```bash
devin mcp list
```

### Use the Spiral skill

1. **Send Windsurf this message**: Your agent reads Spiral's skill and sets itself up.

```
set up https://spiral.ai/SKILL.md
```

2. **Sign in or paste a key**: Approve Spiral in the browser, or give the agent a key from the console.

### Use the Spiral CLI

1. **Install and sign in**: Installs the Spiral CLI, signs you in and adds its skill to your coding agents.

```bash
npx -y spiral-cli@latest init
```

### Call the Spiral API

1. **Create an API key**: In the console, under API keys.

2. **Run Nova-3**: One request; the result comes back by polling or a webhook.

```bash
curl https://api.spiral.ai/v1/runs \
  -H "Authorization: Bearer $SPIRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "provider": "deepgram",
    "model": "nova-3",
    "input": {
      "url": "<url>"
    }
  }'
```

Then paste one of the prompts above: Windsurf finds Nova-3 on Spiral, shows the price and runs it.

Setup guide in the docs: https://spiral.ai/docs/quickstarts/windsurf

## Nova-3 tools and prices

- [Nova-3 Speech to Text](https://spiral.ai/tools/deepgram/nova-3) (Deepgram): $0.00559 per minute of audio, per channel ($0.00676 with language multi). Redaction: $0.0026 per minute. Key terms: $0.00169 per minute. Entity detection: $0.00221 per minute.

## Settings you can ask for

| Setting | Allowed values | Default | Description |
| --- | --- | --- | --- |
| extra |  | — | Arbitrary key-value pairs that are attached to the API response for usage in downstream processing |
| redact |  | — | Redaction removes sensitive information from your transcripts |
| search |  | — | Search for terms or phrases in submitted audio |
| diarize |  | — | Deprecated: use `diarize_model` instead. Recognize speaker changes. Each word in the transcript will be assigned a speaker number starting at 0. Cannot be sent together with `diarize_model`. |
| keyterm |  | — | Key term prompting improves recognition of specialized terminology and brands. |
| replace |  | — | Search for terms or phrases in submitted audio and replaces them |

## What can you do with Nova-3 in Windsurf?

Turn recorded demos, interviews, and technical discussions into project documentation with Nova-3. A coding agent can use vocabulary from the repository and convert timed transcripts into captions or reference notes.

### Meeting notes with traceable quotes

Transcribe a recorded meeting with speaker labels, then have the agent draft decisions and action items with timestamps back to the discussion.

### Interview transcripts

Turn recorded interviews into transcripts organized by speaker. Use timestamps to find and verify quotes before publishing.

### Captions for recorded lessons

Transcribe lesson audio, then have the agent convert the timed words into draft SRT or WebVTT captions for review.

### Multilingual conversation records

Transcribe discussions that switch between supported languages. Have the agent preserve the spoken languages when organizing the transcript.

### Technical training transcripts

Supply vocabulary hints for product names and technical terminology in recorded training sessions. Have the agent use the transcript to draft searchable reference notes.

### Customer call review

Transcribe recorded customer conversations with speaker labels, then have the agent collect relevant quotes and timestamps for a reviewer.

### Recorded lectures from distant microphones

Create draft transcripts of lectures or presentations recorded away from the speaker. Use the timed transcript to check passages against the recording.

## Overview

- [Deepgram](https://spiral.ai/providers/deepgram)
- [All Windsurf guides](https://spiral.ai/integrations/windsurf)

## Nova-3 in other agents

- [Amp](https://spiral.ai/integrations/amp/nova-3)
- [ChatGPT](https://spiral.ai/integrations/chatgpt/nova-3)
- [Claude](https://spiral.ai/integrations/claude-desktop/nova-3)
- [Claude Code](https://spiral.ai/integrations/claude-code/nova-3)
- [Cline](https://spiral.ai/integrations/cline/nova-3)
- [Codex](https://spiral.ai/integrations/codex/nova-3)
- [Cursor](https://spiral.ai/integrations/cursor/nova-3)
- [Devin](https://spiral.ai/integrations/devin/nova-3)
- [dots](https://spiral.ai/integrations/dots/nova-3)
- [Factory Droid](https://spiral.ai/integrations/factory-droid/nova-3)
- [Gemini](https://spiral.ai/integrations/gemini/nova-3)
- [Gemini CLI](https://spiral.ai/integrations/gemini-cli/nova-3)
- [Google Antigravity](https://spiral.ai/integrations/antigravity/nova-3)
- [goose](https://spiral.ai/integrations/goose/nova-3)
- [Grok](https://spiral.ai/integrations/grok/nova-3)
- [Grok Build](https://spiral.ai/integrations/grok-build/nova-3)
- [Hermes](https://spiral.ai/integrations/hermes/nova-3)
- [Instinct](https://spiral.ai/integrations/instinct/nova-3)
- [Manus](https://spiral.ai/integrations/manus/nova-3)
- [Muse](https://spiral.ai/integrations/muse/nova-3)
- [Muse Code](https://spiral.ai/integrations/muse-code/nova-3)
- [n8n](https://spiral.ai/integrations/n8n/nova-3)
- [Notion](https://spiral.ai/integrations/notion/nova-3)
- [OpenClaw](https://spiral.ai/integrations/openclaw/nova-3)
- [opencode](https://spiral.ai/integrations/opencode/nova-3)
- [Perplexity](https://spiral.ai/integrations/perplexity/nova-3)
- [Raycast](https://spiral.ai/integrations/raycast/nova-3)
- [VS Code](https://spiral.ai/integrations/vscode/nova-3)
- [Zapier](https://spiral.ai/integrations/zapier/nova-3)

## More Windsurf guides

- [Mureka 9.5 Instrumental](https://spiral.ai/integrations/windsurf/mureka-9-5-instrumental)
- [Nova-3 Multilingual](https://spiral.ai/integrations/windsurf/nova-3-multilingual)
- [Mureka 9.5](https://spiral.ai/integrations/windsurf/mureka-9-5)
- [Orpheus TTS](https://spiral.ai/integrations/windsurf/orpheus-tts)
- [Mirelo SFX 2.0 Video to Audio](https://spiral.ai/integrations/windsurf/mirelo-sfx-2-0-video-to-audio)
- [Qwen Audio 3.0 TTS Flash](https://spiral.ai/integrations/windsurf/qwen-audio-3-0-tts-flash)

## Questions about Nova-3 in Windsurf

**Can Windsurf use Nova-3?**

Yes. Add Spiral to Windsurf as an MCP server, and Windsurf can find and run Nova-3 like any built-in tool.

**How much does Nova-3 cost in Windsurf?**

$0.00559 per minute of audio, per channel ($0.00676 with language multi). Redaction: $0.0026 per minute. Key terms: $0.00169 per minute. Entity detection: $0.00221 per minute. You pay from your Spiral balance as you go, with no subscription, and a failed run is not charged. Your Windsurf plan is separate.

**Do I need a Deepgram account to use Nova-3 in Windsurf?**

No. Spiral runs Nova-3 on its own Deepgram account. You sign in to Spiral once and never paste a key.

**How long does it take to set up Nova-3 in Windsurf?**

About two minutes: add the server, then sign in, then check it.

**Does Windsurf wait while Nova-3 runs?**

Nova-3 runs in the background. Windsurf starts the run, checks on it and gives you the json when it's ready.

**What is Nova-3 useful for?**

Nova-3 is useful for meeting transcripts, interviews, captions, and recordings with multiple speakers. Deepgram also recommends it for noisy audio and speech recorded far from the microphone.

**Can Nova-3 handle speakers switching languages?**

Yes. Multilingual transcription supports conversations that switch between supported languages, with 55 languages documented by Deepgram.

**How can I improve recognition of technical terms?**

Supply a focused list of vocabulary hints for names, brands, and specialized terminology. Deepgram recommends the most important 20 to 50 terms, within a combined limit of 500 tokens.

**Can Nova-3 separate speakers?**

Speaker diarization detects changes in who is speaking and attaches a speaker label to each word. This helps an agent organize interviews or meeting transcripts by speaker.

**How large can a prerecorded recording be?**

Deepgram documents a maximum file size of 2 GB. For large videos, extract the audio before transcription.

