I have a backlog of 6-hour seminar recordings I need to turn into text. I tried running them through the usual suspects (whispr and some online tools), and they all choke.
Either they hallucinate after 45 minutes, or they hit a file size limit that’s laughably small (like 500mb). It feels like these trillion-dollar companies are intentionally nerfing their tools to force enterprise sales.
I eventually had to find a smaller wrapper tool just to handle a 10-hour audio file without crashing. It’s wild that the "cutting edge" can't handle a simple long-form wav file in 2025.
Is this a context window issue or just lazy product design?
[link] [comments]