# Changelog of Audio & Video Transcriber: Podcasts, SRT, Speakers (`offerastudio/audio-video-transcriber`) Actor

- **URL**: https://apify.com/offerastudio/audio-video-transcriber/changelog.md
- **Full Actor documentation**: https://apify.com/offerastudio/audio-video-transcriber.md

## Changelog

### 1.0.0 (2026-10-07)

- Initial Audio & Video Transcriber Actor.
- Sources: file upload, direct media URLs, podcast RSS feeds (newest episodes first, date and title filters), Apple Podcasts and Spotify links (resolved to the public RSS feed), Google Drive / Dropbox / OneDrive / SharePoint share links, the media URLs in another Apify tool's dataset, and platform links (TikTok, X/Twitter, Facebook, SoundCloud, Dailymotion, YouTube, Reddit, Instagram, Vimeo) through yt-dlp with browser impersonation (audio only).
- Two modes: BYOK (your own Deepgram, AssemblyAI or OpenAI-compatible key, e.g. Groq or OpenAI; every plan) and managed (the owner's keys; paid plans). Managed tries Deepgram nova-3, then AssemblyAI, then Groq whisper-large-v3-turbo, and moves on when a provider's credit runs out.
- Outputs: dataset row with timestamped segments, speaker labels, detected language and optional word timestamps; SRT, WebVTT, text and JSON files in the key-value store.
- Pay per event, by the second: `transcript-second-managed`, `transcript-second-byok` and `platform-proxy-second` (only when a blocked platform link had to go through the residential proxy, paid plans only, one attempt). Failed, empty and info rows are free.
- A watchlist (`watchlistId` + `newItemsOnly`) so scheduled runs only transcribe new media and never charge the same media twice.
- SSRF protection for every download and redirect, retries on 429 and 5xx, sequential processing, partial success.
- Docker image on Alpine with ffmpeg, ffprobe, yt-dlp and curl-cffi.