# Everything AssemblyAI Released in October – 90 Second Update

## Метаданные

- **Канал:** AssemblyAI
- **YouTube:** https://www.youtube.com/watch?v=cLXqHPMJTz4
- **Дата:** 03.11.2025
- **Длительность:** 1:40
- **Просмотры:** 471
- **Источник:** https://ekstraktznaniy.ru/video/12282

## Описание

From Universal Multilingual Streaming to LLM Gateway, we've been busy making speech AI faster, safer, and easier to build with.

In this video, I cover:
- Universal Multilingual Streaming – Real-time transcription in 6 languages with the same speed and accuracy you love
- Speech Understanding – Simplified workflows for speaker ID, formatting, and translation in one request
- AI Safety Guardrails – Profanity filtering, PII redaction, content moderation, and speech thresholds
- LLM Gateway – Access to all major LLMs through one interface (no more juggling multiple accounts!)
- Model Improvements – Better accuracy, automatic code switching, enhanced speaker labels, and more

Key Takeaways:
✅ Universal Multilingual Streaming now supports Spanish, French, German, Italian, and Portuguese
✅ Speech Understanding eliminates multi-step processes for common tasks
✅ New guardrails help you ship safer voice AI apps
✅ LLM Gateway simplifies A/B testing and model access across providers
✅ Universal m

## Транскрипт

### Segment 1 (00:00 - 01:00) []

Hey everyone, Mart here from Assembly AI. I'm going to try and recap everything we released in October in under 90 seconds. First up is universal multilingual streaming. Love our streaming models speed and accuracy. Well, we're bringing that same speed and accuracy to five other languages, Spanish, French, German, Italian, and Portuguese. You can give it a try by setting your speech model to universal streaming multilingual. Next up, speech understanding. We've made it easier for you to do tasks like speaker identification, custom formatting and translation all in one request. There's no need for multi-step processes or an endless refining of prompts. Just use speech understanding and we take care of all of that for you. If you're shipping voice AI apps to users, you want to make sure that your final outputs are safe. That's why we've released guardrails like profanity filtering, PII reduction, content moderation, speech threshold to keep your customers happy and safe while using your apps. We also release LLM gateway which gives you access to all the latest and greatest large language models via one interface. We make it super easy to AV test during development and you don't need one account for every provider anymore, which accelerates your development. Last but not least, we ship model improvements to Universal and Snap. Universal now supports automatic code switching and 200 key terms. Slam is 57% more accurate on critical terms and has 1,000word context aware prompting. We also have 64% fewer speaker counting errors with our improved speaker labels model. That's all we had for this month. Let me know what your favorite release was and if you have any questions, don't be afraid to reach out to Ali. See you.