# Changelog of Audio & Video Transcription: Whisper, SRT & RAG Chunks (`mmdincer/audio-video-transcriber`) Actor

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

## Changelog

### 1.0.6 (2026-11-04)

- Optional glossary (`vocabulary`) and a one-sentence context (`initialPrompt`) are passed to the speech model as its initial prompt, so product names and acronyms are less often spelled phonetically. Same minute events, same prices. Empty by default, so a run without them is unchanged.

### 1.0.5 (2026-11-04)

- Price before publication: Small model per started minute lowered from $0.022 to $0.016 on the Free plan (Bronze $0.0152, Silver $0.0142, Gold and above $0.013). Base and Tiny unchanged at $0.006.
- Documentation: shorter answer at the top of the README, measured price comparison with other transcription Actors (this Actor is not the cheapest), and cross-links to the PDF and video Actors as one RAG family. No code change.

### 1.0 (2026-11-03)

- First release: audio and video files and podcast RSS feeds to timestamped transcripts, SRT/WebVTT subtitles and token-counted chunks with start and end times.
- Whisper speech-to-text (faster-whisper, int8, CPU) with the Small, Base and Tiny models baked into the image. About 100 languages, automatic language detection.
- Pay per started minute: Small model at the primary event, Base and Tiny at a lower fast event. Silence, failed downloads, unreadable files and files over the length limit are free.
- Reads media links from another Actor's dataset (`datasetId`), with limited-permission access.
- `onlyNewEpisodes` remembers transcribed links in a named key-value store of the user's account.
- Safety: private and internal network addresses are refused, 401/403/429 are reported as `blocked` and never worked around, feeds are parsed with entity expansion disabled.