Release Notes

Manual Mode pauses your pipeline at the checkpoints you choose, so you can confirm, modify, or regenerate each intermedi

Blank page after v5.7.3 upgrade fixed — during the merge of the analytics changes into master, git auto-merge treated th

Mode-level analytics for generated videos — the task_completed and task_failed GA4 events now include the mode parameter

Intermediate artifacts restored in the progress panel — after the v5.7.0 frontend re-architecture, intermediate artifact

npm publish step fails with printf: --: invalid option — the release workflow's "Prepare npm README" step used printf '-

Multi-API-Key rotation & throughput boost — configure multiple Agnes API keys (env + config, auto-dedup); requests rotat

Light/dark dual mode — the Midnight design system now supports both light and dark themes with a one-click toggle.

Midnight design system — complete frontend redesign with an atmospheric dark theme, refined typography, spacing and visu

Referral links in the UI and scripts now point to the current official-site paths (deprecated slugs removed).

22-language UI — language support extended to 22 languages, aligned with the official site.

Prompt examples in task forms — each task-type form (Simple / Creative / Manuscript / Anchor) shows a guide link to prom

web/ routing layer — server.py is split into a modular route layer (web/routes/), improving structure and testability.

Optional GA4 analytics — track user operations and error reporting; a setup script auto-configures custom dimensions and

Agnes domain config (com/cn) — configure the Agnes API base domain; switching to the .cn domain now works without the pr

imageio-ffmpeg dependency explicitly declared; npm README documents ffmpeg setup.

imageio-ffmpeg is now an explicit dependency, and the npm README documents ffmpeg setup so video concatenation works aft

Quick Start on Docker / npm pages — Docker Hub overview and npm README both lead with a Quick Start (pull / run / npx fr

Docker Hub overview update now respects the 100-byte short-description limit, so the page title and description publish

Registry page titles (GHCR / Docker Hub) renamed to Free Short Video for consistent product branding.

Release workflow: npm publish is now guarded inside the step with a shell check; missing NPM_TOKEN skips only the npm pu

Release workflow: npm version is synced via npm pkg set (no git commit on detached HEAD), keeping npx free-short-video v

Word-level subtitle timeline (v2.0) — subtitles are aligned to edge_tts word-level cues, the exact source of truth; the

Subtitles in all languages are now capped at two lines for consistent rendering.

Model selector — choose text / image / video models directly in the UI; the new agnes-2.5-flash model shows an in-testin

Docker Hub page shows a concise description and a full README-based overview, refreshed automatically on each release.

Docker Hub overview update now respects the 100-byte short-description API limit, so the description and README publish

GHCR pull path corrected to ghcr.io/lcy362/free-short-video:<version>; the invalid orphan-cleanup step is removed.

Release workflow: orphan-cleanup step no longer references secrets in an if condition, keeping the multi-arch image publ

GHCR images are now pushed with the GITHUB_TOKEN, linking the package to this repository and enabling the multi-arch pul

Docker data persistence — the image now declares VOLUME for /app/.working_dir and /app/.agnes_config; one-command launch

Release notes now show both GHCR and Docker Hub pull commands; the GHCR tag format is corrected to drop the v prefix (e.

Docker builds now default to the official PyPI source; builds inside mainland China can opt into a local mirror via a bu

Resource navigation bar — quick-access links (Demo, Home, Guides, FAQ, GitHub) below the header.

Poetry Video task type — new POST /api/tasks/poetry: the LLM splits a poem into scenes (original line | visual descript

Intermediate-artifacts viewer — the task detail panel shows every step's intermediate outputs (prompts / images / videos

6 new UI languages — German (Deutsch), French (Français), Dutch (Nederlands), Spanish (Español), Portuguese (Português)

Video watermark — automatic watermark overlay on generated videos, with language detection and a UI toggle.

Structured scene configuration — replace the free-text user_requirement with explicit parameters: duration_source (manua

Bilingual screenwriter — all 14 screenwriter meta-prompts support Chinese/English switching.

Multi-workspace management — add, switch, and manage multiple working directories from the Web UI; each workspace keeps

Multi-task concurrency control — a weighted semaphore lets multiple tasks run concurrently with automatic queueing; new

Subtitle & narrator independent control — SubtitleConfig elevated to a peer of AudioConfig, each with its own enabled to

i2i end frame pipeline — image model unified to agnes-image-2.1-flash; i2i array API; character reference images normali

API Key clear/reset — cleanly clear a key in the Web UI instead of only overwriting it.

Three task types with shared foundations — Simple Video (single prompt → single video via Agnes Video API, t2v/i2v/ti2vi

Web UI — one-click launch, operate entirely in the browser, no command line needed.