SubFlow
SubFlow is a professional-grade, privacy-centric subtitle studio that allows creators to translate, edit, and refine video subtitles using local AI processing.
Key Features
100% Private & Local Processing
Video files and data never leave the user's device. All AI processing happens directly within the browser, ensuring absolute security for sensitive content.
Advanced AI Integration
Users can plug in their own API keys for top-tier models like Google Gemini, OpenAI, or Anthropic to generate highly accurate, context-aware translations in multiple languages.
Real-Time Workflow
Offers instant subtitle previews over the video, smart refinement tools (to rephrase or shorten lines), and instant export to industry-standard formats like .SRT.
Open Source Contributions
We are committed to advancing AI for the Portuguese language. Our team has built and published several key datasets, including 'portuguese-ocr-dataset' and 'portugues_ocr_dataset_full', which are widely used by the community to train AI models for reading Portuguese text.
Model Fine-Tuning
To improve document digitization, we actively fine-tune TrOCR (Transformer-based Optical Character Recognition) models. We have released multiple versions of 0.3B parameter models, specifically optimized to accurately extract and read Portuguese text from images.
Applied AI Research
We rigorously experiment with cutting-edge technologies. Our 'Spaces' such as Alpaca-LoRA and Banks demonstrate our expertise in fine-tuning Large Language Models (LLMs) using Low-Rank Adaptation (LoRA) and building robust AI applications.
© 2026 Rua das Pedras Negras, n.º 1, 1.º, dt.º, 1100-404 Lisbon, Portugal.