API under test: convertintomp4.com production API — POST https://api.convertintomp4.com/v1/convert/base64.
Wall time: t0 = immediately after POST returns (upload time excluded); t1 = first poll cycle showing status=completed. Granularity is 3s poll interval, so wallSeconds is quantized at 3s steps. Use processingMs (result.duration from status response) for precise server-side timing.
Inputs: each case used one representative file per format — a single 30-second 1080p clip for video formats, a 60-second WAV/MP3 for audio, a 2000×1500 image, and a synthesized 10-page PDF or DOCX for documents. File details are listed in the downloadable dataset.
Limitations:- Single run, not averaged — timing reflects one execution on the live production server under real concurrent load; it is not a controlled laboratory benchmark.
- One file size per case — results scale with resolution, duration, and bitrate; the numbers here are indicative, not universal.
- No competitor lanes — this study measures only our own API; no third-party services were tested.
- Poll-quantized wall time — wall seconds are quantized at 3 s poll intervals; use
processingMs in the dataset for precise server-side timing. - pdf-to-jpg returns a ZIP — multi-page PDFs produce one JPEG per page packed into a ZIP archive, so the output size reflects all 10 pages.