This issue often occurs with very large projects containing hundreds of slides where audio has been recorded as a single file.  Basically, each PC and browser will handle the rendering and playing of the file slightly differently and at different speeds.

The best way to have your slides synched up nicely with audio is to make sure the voiceover is NOT loading as one huge audio file.  If the person that originally captured all of this did it as one big audio file, then you need to force Captivate to split the audio for each slide.  That way, a separate small audio file is loaded for each slide and the synchronisation is far more reliable.

If you currently have slide audio as one big file with markers set throughout the audio to synch with slides, go into each slide, select the audio waveform on the timeline for that slide and move the audio clip to the right slightly from the beginning of the slide. Do the same at the end of the slide.  The audio file should not touch the beginning or the end of the timeline on the slide. The gap only has to be about one tenth of a second at either end to achieve the result that Captivate will render this chunk of audio out as a separate file at runtime and load it for that particular slide.  Once the audio is broken up like this, the synchronisation should work a lot better.

When you record audio as you capture, this is the way it looks when you go back and look at the slides in Edit mode.

Did you read my instructions about creating a gap at the beginning and end of the slide?  If you don't have this gap present, then when Captivate publishes out the project it will join all the audio files into one big contiguous file that will take longer to load at the begining of the project, and in long projects like yours, can gradually get out of synch with the slides.