Audio Transcriber – Fast Whisper (tiny/base/small) from URL
Pricing
from $12.00 / 1,000 audio minute (small)s
Audio Transcriber – Fast Whisper (tiny/base/small) from URL
Fast, affordable Whisper transcription from direct audio/video URLs (CPU int8). Default small – solid for English & clear speech (DE FLEURS WER ≈10.8%, 60 clips). tiny/base available. Failed files never charged. No YouTube scraping.
Pricing
from $12.00 / 1,000 audio minute (small)s
Rating
0.0
(0)
Developer
Kilian Schieri
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
a day ago
Last modified
Categories
Security
Limited permissionsShare
Audio Transcriber – Fast Whisper from URL
Fast, affordable transcription of audio or video files from direct URLs with faster-whisper on CPU (int8). Choose tiny, base or small (default). Returns plain text, timestamped segments, detected language and duration. The default small model is baked into the image for faster cold starts.
Best for: English and clear, studio-like speech. Measured German WER (own bench, 2026-11-05): small ≈ 10.8 % on FLEURS de_de (60 clips, 1375 words); ≈ 13.6 % on MLS German (60 clips). For higher DE accuracy (~4.7 % FLEURS) use datvanu/audio-transcriber-pro.
No YouTube / TikTok / Instagram page URLs – pass a direct file link (mp3, wav, ogg, m4a, mp4, webm, …). Failed files are ERROR rows and never charged.
Input
| Field | Default | Notes |
|---|---|---|
urls | – | Direct http(s) links, max 50 per run |
model | small | tiny / base / small |
language | auto | ISO-639-1, e.g. en, de |
task | transcribe | or translate (to English) |
includeSegments | true | Timestamped segments[] |
beamSize | 5 | 1 = fastest; up to 5 |
vadFilter | true | Skip long silence |
maxFileBytes | 200 MB | Oversized downloads → ERROR, not charged |
Prefill: short public-domain Aesop fable (~46 s) from LibriVox / Internet Archive.
Output (one dataset item per URL)
status (OK / ERROR), url, model, task, language, languageProbability, durationSecs, audioMinutes (started minutes, ceil), text, segments[] (start, end, text), error, transcribedAt.
Key-value record RUN_SUMMARY holds counts and charged minutes.
Pricing (PPE)
| Event | When |
|---|---|
audio-minute-tiny | started minute with model=tiny |
audio-minute-base | started minute with model=base |
audio-minute-small | started minute with model=small |
apify-actor-start | Actor start (platform standard) |
Minutes are rounded up per file. Failed files are never charged. Remove synthetic apify-default-dataset-item before publish.
Memory
Default 4096 MB. tiny often runs at 1024 MB; small is safer at 2048–4096 MB.
Licence / source notes
- Engine: faster-whisper / CTranslate2, MIT.
- Prefill sample: LibriVox recording of Aesop’s Fables (public domain), hosted on archive.org.
- You are responsible for having the right to transcribe the files you submit (copyright, privacy).
