Comparison of whispers

[JReed]

Hi there. Has anyone compared the whispers? price/speed/quality?
There are many of them on the list, but it’s a pity to waste tokens on testing.

[Muruntaev]

Hey.

I tested some whispers. Threw the results into a flow file. There’s a description of each node’s functions, cost, and execution time.

Flow saved here:
https://drive.google.com/drive/folders/1P5MqqM0Qt0ZJGME0Tn1BIZCPL5BDdPBx?usp=sharing

[JReed]

Thank you. Is there a way to receive the result as a normal text, not in segments?

[Muruntaev]

Haven’t found a built-in solution yet. I have to do it with Python code for each specific node.

Here’s how:

1. Find the “Python code” node and add it to the flow.
   

   

2. Go to the node settings and set the field to “Python code.”
   

   

3. Copy the ID of the whisper node (you’ll need it in the code) and create a link between the nodes.
   

   

4. Now copy this code and replace “NodeID” with the one you copied.

segments = context["NodeID"]["success"]["segments"]
result = ""
for segment in segments:
  result = result + segment["speaker"]
  result = result + ": "
  result = result + segment["text"]
  result = result + "\n\n"
_result = result

If everything worked right, it should look like this:

[JReed]

Thanks for helping me out again. It looks more complicated than I expected. I hope the developers come up with a more convenient way to work with the node data.