Show HN: Free OpenAI Whisper Transcription Tool | Hacker News Reader

Show HN: Free OpenAI Whisper Transcription Tool

whisperapi.com

4 pointsbbakerma5 comments

Hey all! Been working on a transcription API (WhisperAPI.com) powered by the OpenAI Whisper model for a while now. I was getting a good number of requests for a place to drag-and-drop audio files for transcription so people could try out the service. I decided to build just that and make this part of the service public-facing and free. Would love any feedback on it; it’s not as feature-rich as the API itself, but hopefully can be useful for folks.

Open articleView on HN

Interesting. Are there any limits?

And what's your plan with this?

It doesn't seem like you're trying to maximize getting leads by getting mass attention for free (based on the nonexistent copy on the page, with all due respect) so I'm just curious

Currently no per user limits today (other than 60 min per audio file which is built into the API). I may need to add per user limits in the future to be fair to everyone, depending on usage. I initially loaded about 3,000 hours of audio credit into the service to see how that goes; I’m more than happy to increase that if it gets good, organic traffic.

No big plans for the free transcriber at the moment other than to share it with the community and improve SEO.

My goal is to grow the API service and the best way I could think of is to unobtrusively build add on tools and well-researched blog posts to create community value that hopefully drives some API sign ups.

I did my best to add some helpful but not overly sales-y copy to the free speech to text tool page (I targeted about 500 words, most after the tool itself in case people didn’t want to scroll). Any suggestions for copy improvement?

Just be careful to monitor your usage, while it may be free to me, I suspect it is not free to you. Watch for a surprise bill.

Yup, made sure to setup good monitoring/limits, but put a large amount credits in there to make sure it can scale in case it gets good usage :)

> results can be as fast as 10x the audio length

shouldn't this say 1/10th instead?

Reply on news.ycombinator.com