Audio and Video Transcriber: text, SRT and VTT subtitles
Pricing
from $9.00 / 1,000 audio minute transcribeds
Transcribes audio and video files you link to or upload, and podcast RSS episodes, into text with timestamps, SRT and VTT subtitles and the detected language. A Whisper speech model runs inside the actor; no outside service is called.
Direct links to audio or video files, up to 1 GB each: MP3, WAV, FLAC, Ogg, Opus, WebM or MKV, with Opus, Vorbis, FLAC, MP3 or PCM audio and AV1, VP8, VP9 or Theora video. Other files get an error item naming their format. Links to pages on video or social platforms such as YouTube or TikTok are refused. Use files you own or are allowed to process.
Upload files from your computer. When the upload dialog asks for a store, choose a permanent or existing store, then pick that same store under "Key-value stores with your files" so the run can read it.
Apify key-value stores that hold your audio or video files. Every record whose key ends in .mp3, .wav, .flac, .ogg, .oga, .opus, .webm, .mkv, .mka or .ogv is transcribed. The run gets read access to these stores only.
Podcast RSS feed URLs. The newest episodes (see the next field) are taken from each feed's audio or video enclosures.
How many of the newest episodes to transcribe from each podcast feed.
Whisper model size. tiny is the fastest, small is the most accurate, base sits between them.
Two letter language code of the speech, such as en, es, de, fr or ja. Leave empty to detect it from the first 30 seconds of speech.
Write the transcript and subtitles in English, whatever language is spoken.
Add every word with its start time, end time and probability to each segment. Takes longer.
Names, brands and jargon in the recording, separated by commas, so the model spells them right. The first 800 characters are used.
Detect the parts with speech first and transcribe only those. Turn off if words go missing from quiet speech.
