Blurt: push-to-talk dictation on the AssemblyAI Dictation API
Read the Portuguese translation →
Author: AssemblyAI · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Blurt: push-to-talk dictation on the AssemblyAI Dictation API · Channel: AssemblyAI · 01:31 · Watch on YouTube
Topics: Enterprise AI applications, AI developer tools, Multimodal models
Summary
Section titled “Summary”Blurt is presented as a push-to-talk dictation app built on the AssemblyAI Dictation API. Its demo shows speech transcribed across English, Spanish, and Hindi, and custom key terms helping it spell names and medicine names accurately. The app uses one dictation API endpoint to transcribe and clean up speech, including removing ums and ahs, with instructions customizable in the same request.
Key points
Section titled “Key points”- Blurt is introduced as a dictation app built on AssemblyAI’s Dictation API.
- The demo shows language switching between English, Spanish, and Hindi.
- The presenter describes the language switch as smooth and low latency.
- Users can define key terms in settings, including names and medicine names.
- The demo shows those custom terms being spelled correctly in dictated text.
- The app can learn key terms as users dictate and add them.
- Blurt uses one dictation API endpoint for its core functionality.
- The API combines transcription and cleanup, including removal of ums and ahs, in one step.
- Users can customize and send instructions in the same API request.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Blurt demo and multilingual dictation | The presenter introduces Blurt and demonstrates dictation with transcription across English, Spanish, and Hindi. |
| 00:31 | Custom key terms | The demo shows how users can add names and medicine names in settings to improve spelling in dictated text. |
| 01:01 | One-step dictation API | The presenter explains that Blurt uses one API endpoint to transcribe speech, remove filler words, and apply custom instructions. |
Quotes
Section titled “Quotes”I’m making a launch video of our first dictation app called Blurt. — 00:00
You can go to the app, go to settings and define the key terms — 00:31
as you dictate more, it it will learn the key terms. — 01:01
we give you all in one step — 01:31
immediately ums and ahs are removed. — 01:31
Concepts and entities
Section titled “Concepts and entities”Blurt, AssemblyAI, Dictation API, Universal 3.5 Pro, Google Doc
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Hey, I’m making a launch video of our first dictation app called Blurt. To show you a demo, I’m going to go to Google Doc and first I’m going to talk in English and you have no español también. I don’t know about Hindi people but and you will see this transcribed English to Spanish to Hindi. The language switch happens smoothly with lowest latency I’ve seen with the support of 18 languages and this app is built on on our universal 3.5 Pro
00:31 flagship model. I’m going to show you one more cool thing. You can go to the app, go to settings and define the key terms such as names like Nguyen, Vietnamese name, Shawn, Maharashtrian and some medicines like Prinivil, Atorvastatin and we’re going to do a magical demo. I’m going to be like, “Hey, prescribe Prinivil to her, Nguyen. Shawn should take probably Minoxidil and Nguyen should take Atorvastatin and I’m just making a video. Don’t prescribe
01:01 anything. So, now you see that names, medical names, her, Nguyen, Shawn, Nguyen absolutely rightly spelled names, rightly spelled medicines all got it right like magic and you know, as you dictate more, it it will learn the key terms. You can add it there. And that’s not it. This is built on our dictation API. So, this app is basically built with one endpoint which is a dictation call. Typically typically we have dictation apps built on two
01:31 steps. Step number one, voice to text then LLM to clean it up like remove the ums, ahs but we give you all in one step so that you could, you know, customize and send your instructions in the same API request. If you notice, “Hi, um ah I don’t know what I’m saying.” You will see immediately ums and ahs are removed. So, that’s what makes our app Blurt special being built on top of our dictation API. We can’t wait to see how you use it and what you build with our dictation API. Thank you.