Hey! so I’m having difficulty compressing my binary data file (a already handbrake compressed file from 382mb to 105mb) to the point where its 25mb so I can feed it into a OpenAI Whisperer node and transcribe it. I did have the youtube transcription but for my specific purposes I’m trying to use whisperer to transcribe the video better as I believe it has better accuracy? I did try the compress node and also tried to make it gzip but it only went down to 98.4mb which isn’t close. What could I do to compress it more within the automation?
Thank you!
You can try the following, with processing in mind:
Extract audio only: Using a custom FFMPEG node (HTTP Request or Code), convert the video to .mp3 or .wav audio, without the video.
Compress audio: Use ffmpeg to lower the bitrate (e.g., 64kbps, sampling at 16kHz or 32kHz). This allows you to meet the limit without losing much precision.
Check size and split if necessary:
- Use a Move Binary Data or Function node to measure the size.
- If it exceeds 25MB, split it into smaller chunks.
- Send each chunk to the Whisper node in a loop (SplitInBatches).
Manage context (optional, but useful): If you chunk, you can pass the partial transcript as a prompt to the next chunk to maintain continuity.
1 Like
This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.