# Whisper (via Groq)

It's similar to the [Whisper](https://docs.boltai.com/docs/plugins/whisper) plugin, but instead of hosting on OpenAI server, it's hosted by Groq. Unlike other AI services, Groq uses The LPU™ Inference Engine that delivers exceptional compute speed, quality, and energy efficiency.

### How to set up the plugin?

Go to Settings > Plugins > Whisper (via Groq). Select tab `Settings` then enter your Groq API key.

### How to use the plugin?

See the usage on the [Whisper plugin page](https://docs.boltai.com/docs/plugins/whisper).

### What are the main differences?

|            | Whisper via OpenAI                                                                                        | Whisper via Groq                                         |
| ---------- | --------------------------------------------------------------------------------------------------------- | -------------------------------------------------------- |
| Model      | whisper-large-v2                                                                                          | whisper-large-v3                                         |
| Rate limit | High ([up to 500 RPM](https://platform.openai.com/docs/guides/rate-limits/usage-tiers?context=tier-five)) | Low ([20 RPM](https://console.groq.com/settings/limits)) |
| Pricing    | [$0.006 / minute](https://openai.com/api/pricing/)                                                        | Free for now                                             |

### Limitations

It has the same limitation as the Whisper plugin: File uploads are currently limited to 25 MB and the following input file types are supported: `mp3`, `mp4`, `mpeg`, `mpga`, `m4a`, `wav`, and `webm`.