Releases: FP-Studio/framepack-studio
Releases · FP-Studio/framepack-studio
0.5.1
Version 0.5.1
What's Changed
- Use input image as first frame in Original videos by @Insult-to-Ninjary in #312
- fix(toolbox): Correct FFmpeg argument order in tb_create_loop by @ai-anchorite in #328
Thanks!
0.5.0
Version 0.5
Features:
- MagCache has been added and is now the default caching mechanism
- Prompt enhancement with IBM's Granite LLM
- Image captioning with Microsoft's Florence2 LLM
- Docker images are built automatically and available at https://hub.docker.com/r/colinurbs/fp-studio
- New (optional) larger latent preview area
- Improved T2V generations when starting from noise which is now the default latent
Additionally we've recently launched a documentation site at https://docs.framepackstudio.com/
Note: Due to the new LLMs used for captioning and prompt enhancement there are new dependencies. The LLMs will also need 6.25GB of storage. The models will be download the first time you use their respective features.
Special Thanks:
@RT_Borg https://github.com/RT-Borg
@TeslaDelMar https://github.com/ayan4m1
@anchorite https://github.com/ai-anchorite
@Xipomus https://github.com/Xipomus
@contrinsan https://www.youtube.com/@dj__grizzly
@code https://github.com/obfuscode
Zehong Ma
0.4.1
Features:
- New setting to automatically clean temp folder when FPS is started
- Generation presets now include every parameter including prompts
- Batch upload interface for adding a batch of starting images and adding the jobs to the queue
- Export added to post-processing for exporting in different file formats
- Join added to post-Processing for joining clips
- Post-processing workflows to automatically apply a sequence of operations
- Batch post-processing for processing multiple video files
- Expand post-processing frame extraction
- Optional streaming mode in Post-processing to allow for larger input videos and preserve system memory
- Improvements to included batch files
- Post-processing presets
Bug Fixes:
- Lora weights are now applied consistently when multiple loras are used
0.4.0
6-10-2025 Version 0.4
Features:
- Video generation types for extending existing videos including Video Extension, Video Extension w/ Endframe and F1 Video Extension
- Post processing toolbox with upscaling, frame interpolation, frame extraction, looping and filters
- Queue improvements including import/export and resumption
- Preset system for saving generation parameters
- Ability to override system prompt
- Custom startup model and presets
- More robust metadata system
- Improved UI
Bug Fixes:
- Parameters not loading from imported metadata
- Issues with the preview windows not updating
- Job cancellation issues
- Issue saving and loading loras when using metadata files
- Error thrown when other files were added to the outputs folder
- Importing json wasn’t selecting the generation type
- Error causing loras not to be selectable if only one was present
- Fixed tabs being hidden on small screens
- Settings auto-save
- Temp folder cleanup