Closed vidhanbagdai closed 3 years ago
Yes the transcription latency is high 😐 !!
Yes I know! It is impossible to showcase this to potential clients.
Does anyone have a sloution for this?
Hi, might I suggest tuning your lambdas- I added memory to the real-time lambdas and see fantastic performance in my environment. Also, 100mbps is a measure of bandwidth, not an indication of network latency..
@cx-cwiser Tried increasing the memory of lambdas, the transcribed text is coming up quicker on screen but still there is a gap of 5-6 seconds between what the customer speaks and the writing of the transcribed text. Any other suggeestioin or perhaps how much memory to assign to each lambda?
Its due to the Comprehend/Transcribe service itself. Deal with it. If you want something fast use WEB Speech API.
You can also try to keep your KVSTranscriber Lambda function warm by following this blog: https://aws.amazon.com/blogs/compute/new-for-aws-lambda-predictable-start-up-times-with-provisioned-concurrency/
You will need to update the KVSTrigger function to call the pre-warmed version of the function
@vidhanbagdai are you still facing this issue?
There is a delay of almost 6-7 seconds between what I speak and the real time transcription that happens, i.e. the text comes up on screen after 6-7 seconds after I finish speaking. The network speed is almost 100 mbps, still this issue occurs.