Bulk Transcription — Upload 50+ Files at Once

VexaScribe supports up to 50 files per batch upload. As soon as the first batch starts processing, you can queue another 50 — so for a backlog of 200+ files, you can submit them across 4 batches in a few minutes total. The constraint is total minutes per month under your plan ($2/mo = 200 min, up to $20/mo = 6000 min), not file count.

If your 100 files total under 6,000 minutes (100 hours), VexaScribe Studio at $20/month covers the whole batch — that works out to roughly $0.20/hour. If you need pure pay-as-you-go API access (no monthly commitment), Deepgram is currently cheapest at ~$0.22/hour. AssemblyAI is $0.12-0.37/min depending on tier. Rev AI is $0.02/min for English. For 100 typical research interviews (~60 min each), the math: Whisper local install = $0 if you have a GPU; VexaScribe Studio = $20; Deepgram API ≈ $22; AssemblyAI ≈ $30-45; Rev AI ≈ $60; Rev human transcription = $4,500. Pick based on whether you want a managed UI or developer pipeline.

VexaScribe processes files in parallel batches, not sequentially. A batch of 50 one-hour files typically completes in 30-60 minutes total (vs ~25 hours if processed one at a time). Each individual file still takes its normal ~5-10 minutes per hour of audio — the parallel processing just runs many files concurrently. Close your browser; you'll get notified when each completes.

Use VexaScribe's web UI when you have a one-time backlog (research interviews, podcast back-catalog, training video library), need to review/edit transcripts before exporting, or your team isn't developer-led. Use a developer API (AssemblyAI, Deepgram, Rev AI, or OpenAI Whisper API) when you need a recurring automated pipeline — for example, automatically transcribing every new episode of a podcast as it publishes, or integrating with an internal CMS. VexaScribe doesn't currently offer a public REST API; for that workflow, we recommend Deepgram or AssemblyAI directly.

Audio: MP3, WAV, M4A, FLAC, OGG, AAC, WMA, AIFF. Video: MP4, MOV, AVI, MKV, WEBM, FLV, WMV. You can mix formats freely in the same batch — process 20 podcast MP3s and 30 webinar MP4s together. Maximum file size is 5 GB per file (most free tools cap at 25 MB). No need to extract audio from video first — we handle it internally.

Yes. After a batch completes, click "Download All" to get a ZIP archive containing every transcript. Pick the format once (TXT, DOCX, SRT, VTT, or PDF) and it applies to all files. Each transcript is named to match the source filename (interview_001.mp3 → interview_001.txt), which keeps research and content workflows organized. For NVivo, Atlas.ti, MAXQDA, and Dovetail pipelines, the TXT or DOCX format imports cleanly.

Yes. Every file in the batch receives speaker diarization independently — "Speaker 1" through "Speaker N" per file. You can rename speakers in each transcript individually after processing (e.g., "Speaker 1" → "Participant A" for research, or → actual interview subject names). Best accuracy is with 2-6 distinct speakers per file; very similar voices or overlapping speech are the main accuracy challenges.

Yes. Mix any of the 99 supported languages in a single batch — 20 Spanish interviews, 15 English customer calls, 10 Mandarin podcast episodes, all in one upload. Each file's language is auto-detected, or you can specify per file for best accuracy. Each transcript outputs in its source language (cross-language translation isn't included — for that, see our translate-while-transcribing tool).

Yes. The same Whisper-based ASR engine processes every file — quality doesn't degrade with batch size. Per-file accuracy depends on audio quality (clean speech: 92-97% word accuracy per the Open ASR Leaderboard; noisy speech with accents: 80-90%). The variation you'll see across a batch reflects the underlying audio variation, not VexaScribe getting tired of processing.

For most business and research bulk uploads (interviews, podcasts, training, sales calls, internal meetings), VexaScribe is appropriate — we don't train models on customer audio, files are encrypted in transit and at rest, and you can delete any file at any time. For genuinely sensitive batches — HR investigation recordings, attorney-client privileged depositions, clinical/therapy sessions, classified content — install OpenAI Whisper locally so the files never leave your computer. For IRB-supervised research, check your IRB's data-residency requirements before uploading externally.