ALMARFELD SubFlow is a free, MIT-licensed Windows desktop workflow for subtitle translation and media-library processing. Discover embedded or adjacent subtitles, choose tracks and target languages, translate, review and write new MKV files with additional subtitle tracks. Original media is not modified; video and audio streams are copied rather than re-encoded.
A resumable batch queue, translation cache, glossary and subtitle editor support collections. Input includes supported text subtitles and PGS image subtitles via Windows OCR. Validation checks translation response structure and output stream metadata; these checks do not certify translation accuracy or hash every media packet. Human review remains necessary.
Local translation uses a separately installed Ollama model by default. Optional compatible API, DeepL and Google Cloud engines require explicit consent before sending subtitle text and may charge separately. Video and audio are not uploaded. FFmpeg and Ollama/models are separate dependencies, with optional MKVToolNix and OCR language components; downloads require a user request. No telemetry; the optional GitHub update check is off by default.
Version 1.2.3 is available for Windows 10/11 x64 as an installer and self-contained portable ZIP. No separate .NET runtime or WebView2 is required. The Windows binaries are not Authenticode signed.
Retains fixes for removal of selected queue items and prevents duplicate entries when adding the same folder again. Removing a waiting item leaves the remaining queue intact; stopping a running item requires confirmation.
Version 1.2.3 adds a Setup page with confirmed dependency installation, model downloads with progress and cancellation, and model selection. Waiting queue items are explained on the Dashboard. Local Ollama output is bounded; truncated responses are rejected and difficult batches retry at smaller sizes. Translation checks remain heuristic and output still needs review.
No comments or reviews, maybe you want to be first?