How to Use Nova-3 Multilingual in Codex
Use Nova-3 Multilingual to transcribe recordings where speakers change languages mid-conversation. It recognizes 55 languages and accepts vocabulary hints for product names, technical terms, and other words a general transcript might miss.
Prompts to try with Nova-3 Multilingual in Codex
Copy an example into Codex, or ask in your own words.
Use Nova-3 Multilingual to transcribe recordings/customer-interview.wav. Build a transcript viewer in this project that jumps to the recording when I select a word.
Transcribe recordings/product-demo.mp3 with Nova-3 Multilingual. Use product names from the README as vocabulary hints, then save the transcript and word timings as a JSON fixture.
Use Nova-3 Multilingual on recordings/bilingual-meeting.wav. Hint English and Spanish, then create a searchable transcript page with timestamp links.
Transcribe recordings/lecture.mp3 with Nova-3 Multilingual. Turn the returned word timings into a draft caption file and flag caption breaks for review.
Use Nova-3 Multilingual to transcribe recordings/support-call.wav. Save the original response, then create a review table containing each quoted complaint and its timestamp.
How to set up Nova-3 Multilingual in Codex
Pick the way that suits you. Each takes about two minutes.
Use Spiral's MCP server
- Add the hosted server
codex mcp add spiral --url https://api.spiral.ai/mcp
- Sign in
Codex opens the browser: sign in to Spiral and click Authorize. No API key. codex mcp list then shows Spiral.
codex mcp login spiral
Use the Spiral skill
- Send Codex this message
Your agent reads Spiral's skill and sets itself up.
set up https://spiral.ai/SKILL.md
- Sign in or paste a key
Approve Spiral in the browser, or give the agent a key from the console.
Use the Spiral CLI
- Install and sign in
Installs the Spiral CLI, signs you in and adds its skill to your coding agents.
npx -y spiral-cli@latest init
Call the Spiral API
- Create an API key
In the console, under API keys.
- Run Nova-3 Multilingual
One request; the result comes back by polling or a webhook.
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-multilingual", "input": { "url": "<url>" } }'
Then paste one of the prompts above: Codex finds Nova-3 Multilingual on Spiral, shows the price and runs it.
Nova-3 Multilingual providers and prices
Spiral runs Nova-3 Multilingual through Deepgram. Open a card for today's price, with every fee included.
$0.00676 per minute of audio, per channel. Redaction: $0.0026 per minute. Key terms: $0.00169 per minute. Entity detection: $0.00221 per minute.
Settings you can ask for
Nova-3 Multilingual works from your URL. Ask Codex for any of these in your prompt.
| Setting | Allowed values | Default | Description |
|---|---|---|---|
| Diarize | on or off | Deprecated: use | |
| Encoding | linear16, flac, mulaw, amr-nb, amr-wb, opus, speex, or g729 | Specify the expected encoding of your submitted audio | |
| Numerals | on or off | off | Numerals converts numbers from written format to numerical format |
| Dictation | on or off | off | Dictation mode for controlling formatting with dictated speech |
| Punctuate | on or off | Add punctuation and capitalization to the transcript Left unset, smart_format adds punctuation; false turns it off even with smart_format. | |
| Utt split | a number | 0.8 | Seconds to wait before detecting a pause between words in submitted audio |
| Paragraphs | on or off | off | Splits audio into paragraphs to improve transcript readability |
| Utterances | on or off | off | Segments speech into meaningful semantic units |
| Filler words | on or off | off | Filler Words can help transcribe interruptions in your audio, like "uh" and "um" |
| Measurements | on or off | off | Spoken measurements will be converted to their corresponding abbreviations |
| Multichannel | on or off | off | Transcribe each audio channel independently |
| Smart format | on or off | off | Apply formatting to transcript output. When set to true, additional formatting will be applied to transcripts to improve readability |
| Diarize model | latest, v1, or v2 | Select and enable a specific diarization model version. | |
| Detect entities | on or off | off | Identifies and extracts key entities from content in submitted audio |
| Profanity filter | on or off | off | Profanity Filter looks for recognized profanity and converts it to the nearest recognized non-profane word or removes it from the transcript completely |
Every API field of Nova-3 Multilingual Speech to Text is on its tool page: open its card above.
What can you do with Nova-3 Multilingual in Codex?
Turn multilingual recordings into searchable project evidence with Nova-3 Multilingual. A coding agent can use the transcript and word timings to build transcript viewers, editing tools, or review workflows.
Bilingual support call review
Use Nova-3 Multilingual to transcribe calls that alternate between English and Spanish. Have your agent collect complaint excerpts with timestamps for a reviewer.
Multilingual meeting records
Transcribe a meeting with Nova-3 Multilingual, keeping language switches in the transcript. Ask your agent to draft decisions and action items with supporting timestamps.
Customer research interviews
Use Nova-3 Multilingual for interviews where customers mix languages and product names. Have your agent organize timestamped quotes by research question.
Podcast transcript editing
Transcribe a multilingual podcast with Nova-3 Multilingual. Use the word timings to prepare an editing draft and locate passages that need a listening check.
Recorded lecture study notes
Use Nova-3 Multilingual to transcribe lectures that mix a supported language with English technical vocabulary. Ask your agent to create study notes linked to the relevant moments.
Retail conversation analysis
Transcribe recorded customer conversations with Nova-3 Multilingual, supplying relevant product names as vocabulary hints. Have your agent gather examples of recurring product questions.
Multilingual voicemail review
Use Nova-3 Multilingual to transcribe voicemail recordings without choosing a language in advance. Ask your agent to draft a callback list from the transcripts.
More on Nova-3 Multilingual and Codex
Overview
Nova-3 Multilingual in other agents
Questions about Nova-3 Multilingual in Codex
Can Codex use Nova-3 Multilingual?
Yes. Add Spiral to Codex as an MCP server, and Codex can find and run Nova-3 Multilingual like any built-in tool.
How much does Nova-3 Multilingual cost in Codex?
$0.00676 per minute of audio, per channel. 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 Codex plan is separate.
Do I need a Deepgram account to use Nova-3 Multilingual in Codex?
No. Spiral runs Nova-3 Multilingual 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 Multilingual in Codex?
About two minutes: add the hosted server, then sign in.
Does Codex wait while Nova-3 Multilingual runs?
Nova-3 Multilingual runs in the background. Codex starts the run, checks on it and gives you the json when it's ready.
Which languages does Nova-3 Multilingual recognize?
Deepgram currently documents 55 languages, including English, Spanish, French, German, Hindi, Japanese, Arabic, and Korean. It can detect the spoken language without a language hint.
Can Nova-3 Multilingual handle speakers switching languages?
Yes. Its transcriptions follow the speaker's language switches, making it useful for bilingual conversations and meetings where people alternate languages.
How should I guide recognition when I know the languages?
Hint the languages you expect to hear. Hints steer detection without preventing Nova-3 Multilingual from transcribing another supported language.
How can I improve recognition of names and technical terms?
Supply relevant product names, proper nouns, and technical phrases as vocabulary hints. Keep the list focused: Deepgram recommends 20 to 50 important terms, below the 500-token limit.
What does Nova-3 Multilingual return?
Recorded transcription returns JSON containing transcript text and word start and end times. Your agent can use those timings to build a transcript with links back to the recording.
Start using Nova-3 Multilingual in Codex
One Spiral account for every tool your agent needs. No subscription; you pay per run.
Paste this into Codex:
set up https://spiral.ai/SKILL.md


