Introducing Perigee-1: The New Standard for African Language Transcription

We are thrilled to announce the release of Perigee-1, our latest and most advanced speech-to-text model designed specifically for African languages. With Perigee-1, we are setting a new benchmark for accuracy, speed, and reliability in transcribing diverse African dialects and accents.

Why Perigee-1?

African languages present unique challenges for AI models due to their tonal nature, rich morphology, and code-switching prevalence. Perigee-1 has been trained on an extensive dataset of high-quality audio across dozens of languages, including Hausa, Yoruba, Igbo, Swahili, and Pidgin.

Key Features

• Superior Accuracy: Significantly lower Word Error Rate (WER) compared to previous models and competitors.

• Enhanced Code-Switching Support: Seamlessly handles mixed-language audio, a common feature in everyday African conversations.

• Cost-Effective: Unbeatable pricing at just $0.21 per hour of audio.

Supported Languages

Perigee-1 is optimized for a wide range of African languages and dialects:

• Hausa (Orange)

• Yoruba (Deep Purple)

• Igbo (Yellow)

• Swahili (Green)

• Pidgin (Grey)

• Kinyarwanda (Red)

• English (Global)

We are continuously adding support for more languages to bridge the digital divide.

Performance Benchmarks

Perigee-1 demonstrates impressive results on standard benchmarks, achieving low Word Error Rates (WER) across multiple datasets:

Head-to-Head Comparisons

We went beyond standard academic benchmarks, pitting Perigee-1 directly against the world’s leading generalist models. The results expose a significant market gap: while global giants excel at English, they struggle to process the complex realities of African languages.

Perigee-1 outperforms industry giants by wide margins on African languages.

Note: Lower WER (Word Error Rate) is better.

The Bottom Line: You no longer have to settle for broken transcripts when working with African audio. While competitors hold their own on English, Perigee-1 is the dominant leader for African linguistic diversity, offering the only commercially viable solution on the market.

Model Selection & Pricing Transparency

With the release of Perigee-1, we are also introducing Model Selection in our SDKs. This gives you the flexibility to choose the right model for your specific use case and budget.

You can now list all available models and their pricing directly from the SDK:

^{import orbitalsai}

^{client = orbitalsai.Client(api_key="your_api_key_here")}

^{# List available models}

^{models = client.get_models()}

^{for model in models:}

^{print(f"{model.model_name}: ${model.transcription_rate_per_hour}/hr")}

How to Use Perigee-1

Using Perigee-1 is as simple as specifying the model_name parameter in your transcription request. If you don't specify a model, Perigee-1 is now the default!

Python SDK Example

Ensure you have updated to the latest version of the SDK (pip install --upgrade orbitalsai).

^{import orbitalsai}

^{client = orbitalsai.Client(api_key="your_api_key_here")}

^{transcript = client.transcribe( "interview_recording.mp3", language="swahili", model_name="Perigee-1" )}

^{print(f"Transcription: {transcript.text}")}

Async Support

For high-volume workloads, our async client also supports model selection:

^{import asyncio}

^{import orbitalsai}

^{async def main():}

^{async with orbitalsai.AsyncClient(api_key="your_api_key_here") as client:}

^{transcript = await client.transcribe( "podcast_episode.mp3", model_name="Perigee-1" )}

^{print(transcript.text)}

^{asyncio.run(main())}

Get Started Today

Perigee-1 is available now for all users. Log in to your dashboard to get your API key and start building the next generation of voice-enabled applications for Africa.