OpenAI charges by the minute, so speed up your audio
10 months ago
- #Transcription
- #Cost-Saving
- #OpenAI
- Speeding up audio before transcription with OpenAI can save time and money.
- Using ffmpeg to increase audio speed (2x or 3x) reduces token usage without significant quality loss.
- OpenAI charges based on audio duration or tokens, making faster audio cheaper to transcribe.
- The gpt-4o-transcribe model has a 25-minute limit, but speeding up audio can help fit longer files.
- 2x and 3x speeds maintain transcription accuracy, while 4x becomes unusable.
- Cost savings can be up to 33% by using 3x speed compared to original audio.
- Output tokens remain consistent across different speeds, suggesting the model compensates for faster input.
- This method is a simple, effective hack for reducing transcription costs.