# Changelog of Audio & Video to Text Transcription (Whisper): Speech to Text (`fguiraud/audio-video-transcriber`) Actor

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

## Changelog

User-facing changes. Dates are UTC.

#### 2026-10-04

- AI agents: common field names (`url`, `urls`, `startUrls`, `query`, `keywords`...) are accepted, and an empty input runs a small sample from the input form instead of failing.

#### 2026-10-02

- Runs now finish before their timeout instead of being stopped by it: a long file is transcribed up to what the remaining time allows (with a warning), and files that no longer fit are marked `skipped` and not charged.

#### 2026-10-01

- TikTok, Instagram, X and Facebook links get a clear message pointing to Video to Text Transcriber, and are not billed.

#### 2026-09-30

- Music or silence that Whisper fills with stock phrases ("you", "thank you") now counts as no speech and is not billed.
- Billed minutes are rounded with a 2-second tolerance: a 60.02-second clip is billed as 1 minute, not 2.
- Fixed: "Only new episodes" could stop the run in accounts that used it before; it now never fails the run.

#### 2026-09-29

- README: measured accuracy (word error rate) and speed of the tiny, base and small models.

#### 2026-09-28

- Readable subtitles: `subtitleMaxChars`, `subtitleMaxLines` and `subtitleMaxDuration` re-cut SRT/VTT into broadcast-style lines.

#### 2026-09-26

- Fixed: rare runs failed with "Schema validation failed" when a result had empty fields.

#### 2026-09-23

- First public version.