# Changelog of Audio & Video Transcription to JSON: RAG Chunks + Timestamps (`landoz/transcribe-to-json-chunks`) Actor

- **URL**: https://apify.com/landoz/transcribe-to-json-chunks/changelog.md
- **Full Actor documentation**: https://apify.com/landoz/transcribe-to-json-chunks.md

## Changelog

### 0.1.6 (2026-10-23)

- Pricing set from measured platform cost: standard $0.012 per audio minute, accurate $0.035 per audio minute.
- Default memory raised to 4096 MB (about one CPU) and default timeout to 4 hours for long recordings.

### 0.1 (2026-10-23)

- First release: audio and video URLs in, structured JSON out.
- RAG chunks in the family chunk schema shared with pdf-to-json-tables (chunkIndex, text, charStart, charEnd, citation), with time citations (startMs, endMs, label like 00:12:03-00:12:48) instead of page citations.
- Timestamped segments with confidence, optional word timestamps, full text, SRT and VTT subtitle files in the key-value store.
- Two accuracy tiers: standard (Whisper base) and accurate (Whisper small), both open-source whisper.cpp on CPU, multilingual with automatic language detection and optional translation to English.
- Any container ffmpeg can read: MP3, WAV, FLAC, OGG, Opus, M4A, AAC, WMA, AIFF, MP4, MOV, WebM, MKV, AVI.
- Pay-per-event pricing: one audio-minute-standard or audio-minute-accurate event per started minute, charged after the duration is known and before transcription; failed downloads and undecodable files are never charged.