Support

Facebook and YouTube occasionally serve a page variant our extractor cannot parse. Please wait a moment and try again — this usually resolves on its own. If the same URL keeps failing over an extended period, the video may be private, region-locked, or a live stream, which we cannot transcribe.

Transcripts come from machine speech recognition (Whisper). Quality depends heavily on audio clarity, background noise, accents, and how much code-switching is in the audio. The accuracy badge on each transcript reflects the model's own confidence — a rough indicator, not a guarantee.

Every transcript runs through up to three auto-escalating tiers (small Whisper → medium with denoise → OpenAI large) — the system automatically reaches for the heavier engines when a tier comes back below its expected band, at no extra charge. See /sla for the full accuracy bands per audio type and what affects accuracy.

If a transcript is still low quality after the auto-ladder, click Re-transcribe + denoise on its row in /dashboard — it forces a fresh run through the higher tiers (denoise + larger model + song-mode params) without re-charging.

Free minutes reset on the 1st of every month — 300 minutes total, with a 60-minute daily cap and a 20-minute per-file cap to keep things fair. The paywall modal shows the exact reset date when you hit a limit. If you don't want to wait, topping up $5 gives you wallet credit that doesn't expire.

The free tier is shared across all browsers on the same network and device, so if someone else on your Wi-Fi has used it, the counter reflects that. The 60-minute daily cap and 20-minute per-file cap also kick in separately from the monthly pool — those can block a single heavy session even if your monthly budget still has minutes left. Topping up removes the limits entirely.

If a paid transcription failed and you weren't able to download the result, we'll refund the full charge. Email us with your Stripe receipt number. We don't refund unused wallet balance because it doesn't expire — you can use it any time.

Currently only YouTube and Facebook URLs are guaranteed to work. Other platforms may work if their content is reachable by yt-dlp, but we don't officially support them. Email us if there's a platform you need.

Up to 4 hours per file. Longer videos take several minutes to process and the price scales linearly at 6¢/minute. For anything longer, split it into shorter segments.

Free transcripts are not stored on disk after the session ends. Paid transcripts are stored long enough to deliver your download link (currently 1 hour) and then deleted. We never train models on your audio or transcript data.