Roadmap

3.0.0

• Introduce pixel boost to the face swappers

• Multi angle enhanced face detector and face landmarker

• Release the GHOST face swapper models to the public

• Update to Gradio 4

• Official support for CUDA 12 and OpenVino 2024

• Return correct error codes in headless mode

• Drop generating output paths to allow automated jobs

• Deprecate config, source, target and output arguments

• Add multi-range slider for trim frames

• Introduce --output-audio-encoder argument

• Transfer order, age and gender to the face selector

• Fix image quality for webp images

• Fix the termination of ffmpeg processing

• Drop the face detector model yunet due low accuracy

• Replace globals with context aware state manager

• Release frame enhancer real_esrgan_x8 models to the public

• Introduce the age modifier processor

• Introduce jobs to support queued processing

• Reduce the thread count and queue count to prevent false usage

• Significantly more accurate benchmark results

• Introduce commands to the CLI

• Rename frame processor to processors

• Fix threading and preview crashes for ROCm

• Live Portrait powered expression restorer processor

• Live Portrait powered face editor processor

3.1.0

• Early Access: cloud processing of remote jobs

• Support audio to image for the lip syncer

• Introduce face animator for static images

• Introduce the hair modifier frame processors

• Introduce the face toonifier frame processors

• Upload and download job assets via --job-upload and --job-download

• Optimize existing jobs via --job-optimize

• Detect webcam devices via ffmpeg to make them selectable

• Expose benchmark and webcam to the CLI

• Add different face parser models like FaRL