4 articles about transcription.
Going from a transcription API to SRT is where caption pipelines break. Fix timestamps, line wraps, and escaping, then burn them on with FFmpeg or one call.
n8n large video transcription fails because n8n holds the whole file in memory. Hand the Drive URL to an audio-extraction API and send only a 4 MB track.
The FFmpeg 16kHz mono WAV command Whisper needs, plus the pipeline half: pulling audio from an MP4, size math, channel picks, and no local binary to install.
The Whisper 25MB limit is a byte cap, not a time cap. Re-encode first, split only when you must, and re-offset every timestamp so your SRT still lines up.