Add output format options

[zappityzap]

• Add checkbox to save individual frames
• Add checkbox to save GIF
• Add checkbox to save MP4
• Move location of saved frames to dated folders under AnimateDiff instead of regular txt2img folder
• Don't save frames by default
• Save GIF by default

Saving as WEBP and WEBM would be nice, but I couldn't get it working.

The UI was rearranged slightly as well. Here it is before:

And here it is after:

[zappityzap]

• Adds text output format with generation parameters, disabled by default, closes [Feature]: Add an option to generate a text file next to each gif, containing the generations parameters  #55
• Adds option to optimize GIFs with gifsicle, disabled by default
  • To use optimize GIF: apt install gifsicle, then activate the venv, pip install pygifsicle
  • Optimized GIFs are smaller and more compatible, but may have less quality.
• Saves to GIF, MP4 and PNG by default
• Uses mm_sd_v14.ckpt by default
• Saves each output format to separate sub-directory
• Fixes UI for various size screens
  

[christopherwolfsaal]

When can this PR be merged? I need the mp4 output

[zappityzap]

Maintainer is busy with real life stuff. You can install it from my fork on the feature-output-formats branch.

I'll be submitting similar PRs to this newer fork: https://github.com/JaredTherriault/sd-webui-animatediff

[continue-revolution]

I will merge, but will do this after some other update in this extension and some fucking real life stuff

[FizzleDorf]

I think it should be rekitted to create a .bat with the commands, place it in the folder, and run it so the user can switch framerates and add filters after the fact. You can even include the missing formats using ffmpeg instead

[zappityzap]

That's a bit out-of-scope for the extension, and it would need to support Linux as well. Saving PNG frames should allow for a user to create their own script.

@continue-revolution Please feel free to pick and choose anything out of this PR and then close it, no worries.

[christopherwolfsaal]

I already implemented a local mp4 processing and making the frames smooth with interpolate..maybe you want to integrate that

def postprocess(self, p: StableDiffusionProcessing, res: Processed, enable_animatediff=False, loop_number=0, video_length=16, fps=8, model="mm_sd_v15.ckpt"):
    if enable_animatediff:
        self.restore_ddim_alpha(p)
        self.remove_motion_modules(p)
        video_paths = []
        self.logger.info("Merging images into MP4.")
        from pathlib import Path
        Path(f"{p.outpath_samples}/AnimateDiff").mkdir(exist_ok=True, parents=True)                   
        dirname=f"{p.outpath_samples}/AnimateDiff"                              
        for i in range(res.index_of_first_image, len(res.images), video_length):  
            video_list = res.images[i:i+video_length]
            seq = images.get_next_sequence_number(dirname, "")
            filename = f"{seq:05}-{res.seed}"
            mp4_path = os.path.join(p.outpath_samples, "..", "mp4", dirname)
            os.makedirs(mp4_path, exist_ok=True)                
            video_pathtemp = os.path.join(mp4_path, f"{filename}_tmp.mp4")
            video_path = os.path.join(mp4_path, f"{filename}.mp4")
            video_paths.append(video_path)
            imageio.mimsave(video_pathtemp, video_list, fps=fps)      

            ffmpeg_command = [
                'ffmpeg',
                '-y',
                '-i',
                video_pathtemp,
                '-crf',
                '10',
                '-vf',
                'minterpolate=fps=60:mi_mode=mci:mc_mode=aobmc:me_mode=bidir:vsbmc=1,setpts=2.5*PTS',
                '-map',
                '0',
                '-crf',
                '10',
                '-an',
                '-f',
                'mp4',
                video_path
            ]
            
            # Execute the ffmpeg command
            try:
                subprocess.run(ffmpeg_command, check=True)
                print("FFmpeg command executed successfully")
            except subprocess.CalledProcessError as e:
                print("Error executing FFmpeg command:", e)               
          
        res.images = video_paths
        self.logger.info("AnimateDiff process end.")