Deepgram is listed in Forbes’ top 50 most promising AI companies to watch in 2021!


Quick Guide to the Deepgram Brain Transcription API

BY Scott Stephenson  | March 08 2018

This is a quick guide for using Deepgram Brain, our general transcription service.


You can both upload a local file (one that is on the computer running the request) or a remote file that is publicly accessible (e.g. hosted in AWS S3 or another server).

Local Upload

For local upload, let’s say you have a file /path/to/file.mp3 on your computer (you can scroll down for similar directions working with files hosted on a server). You can upload this mp3 file to Brain using the following curl command (filling in USER and PASS with your username/password or API ID/token):

curl -u USER:PASS -X POST -T /path/to/file.mp3 -H 'Content-type: application/octet-stream'  

Wait a little bit and you’ll get a JSON response like this:

    "content_url": "...",
    "confidence": [0.999649, 0.995753, ...],
    "duration": 21.43,
    "asset_id": "05e36324fca6c919",
    "transcript": "Hello and how are ...",

If you want to re-download this JSON blob at a later point, you just need to use the asset_id, like so:

curl -u USER:PASS  

Additionally, to get word-level timings for each word, add the ?times=true query parameter:

curl -u USER:PASS  

To list all your available assets, you can simply do:

curl -u USER:PASS  

This sort of “global” listing of all assets doesn’t contain quite as much information as a specific asset query, so if you want more information about a particular asset, just query for the asset directly, as in the previous examples (.../assets/ASSET_ID).

If you want to queue an asset for processing, but don’t want to wait for it to complete, you can simply add the async=true query parameter, like so:

curl -u USER:PASS -X POST -T /path/to/file.mp3 -H 'Content-type: application/octet-stream'  

Remote Upload

If the media file you want transcribed is already available from a public URL, you can avoid the overhead of a file upload by submitting the URL directly to Brain:

curl -u USER:PASS -d '{"audio": {"url": ""}, "config": {"async": true}}'  

Note that in this case, where we submit a JSON payload, the async flag goes in the config field of the JSON object (if the config field is absent, async defaults to false). With this type of request, the file will be downloaded by the Deepgram Brain API then processed in the standard way.

Finally, a reason to get excited about voice recognition.

Check out Deepgram

Contact Us

Speech recognition is hard. We'll make it easy.