中文 Back to home

Changelog

Everything shipped, newest first · October 15, 2026

Public changelog for the Voxtakes dubbing studio. Beta velocity means dates occasionally compress into one entry; version numbers quoted are the render-worker builds shipped to production.
October 15, 2026

Open beta, and the studio speaks Chinese now NEW

  • Open beta: sign-up without invite codes, a short application, human review, 5 free minutes on approval.
  • Bilingual app UI: the whole studio (upload, review, history, billing, terms pages) switches between English and Chinese; errors are localized too.
  • Brand glossary: lock terms verbatim or pin official renderings per language, with a per-line enforcement re-check that flags any term that slipped. One-click "keep original" from any review line.
  • Burned-in subtitle removal: OCR + inpaint erases hardcoded captions inside the pipeline before dubbing (+0.5 credit/min, once per job, reused across languages and redubs).
  • Line review upgrades: rounded durations, sticky approve bar, honest per-QC badges.
October 14, 2026

Delivery modes & caption presets NEW

  • Choose per job: subtitles only (0.4 credit/min), voice dub, or both.
  • Four caption presets that ship burned-in with every language: Clean · Bold Pop · Word Highlight · Bilingual — styling without a separate captioner.
  • Per-line re-dub with A/B: re-record one sentence (text, take, or voice), keep both takes side by side, pay only for the line. QC-caused re-dos are free.
  • Engine routing went live: each language renders through the best-validated voice engine (self-hosted clone + multilingual models), per-line machine QC scores shown in review.
October 13, 2026

Bigger jobs, whole-package bundles IMPROVED

  • Bundle download is one streamed zip: every language's video + SRT/VTT + per-line audio + term report.
  • Long-job ceilings raised (60 min / 2 GB per job verified end-to-end on production).
October 12, 2026

Translation quality step-up IMPROVED

  • Primary translation engine switched to a dedicated MT model with structured fallbacks — brand names, numbers and sentence splitting measurably improved.
  • Credit top-ups landed; ledger now shows every charge with its reason.
  • Background-music separation tightened for dubbing passes (voice removal without losing the bed).
October 8–11, 2026

The dubbing pipeline takes shape NEW

  • Upload-first flow with presigned direct-to-storage upload; URL pull added where platforms allow.
  • Voice cloning from the video itself; multi-speaker keeps interviews as two people.
  • First multilingual end-to-end renders (EN/ES/DE/JA/ZH validation set), full-sentence subtitle segmentation, speed controls per line.
  • Invite-only beta opens to first testers.