Experiments

Nari AI

2025
FINISHED - PASSEDyesAudio Gen

Expressive AI audio with emotion as the differentiator - ran on EC2, documented constraints.

Nari AIEC2Audio Gen

Aim -

try the new Nari AI model.

run it and generate outputs.

document the constraints, limitations and workarounds.

Plan -

  1. join discord
  2. go through docs
  3. run it locally / on ec2 / collab
  4. test variations and document outputs

Implementation -

  1. joined discord. love the outputs people are posting.
  2. skimmed through the doc and implemented this in my EC2.
  3. asked grok to give me few scripts for this, below are my results.

the testing is not completed yet. need to do a deep dive, to find ways to leverage it to its full potential. although I strongly believe that this can help me make engaging yt shorts/reels

Output -

  1. the emotions that you get out of this is the moat.
  2. the script has to be short (less than 20sec). otherwise, it speeds up the narration.
  3. only use tags (non-verbals) listed in the readme. otherwise it goes bonkers (for angry - it added a plane sound in the middle of the conversation).

Next experiment

gpt-image-1 - Image Generation

gpt-image-1 - Image Generation

FINISHED - PASSEDImage Gen