Pipeline


Text

Text

ChatGPT generated the content and storyboard based on the specific conditions that humans restrained. Then, the final version of the story script and storyboard was output after rough edits by humans (such as deleting redundant or unnecessary sentences).

Logo1

Images

Images

Copied the storyboard content into Midjourney as prompts to generate scene and character images. Then, used the "Expand Image", "Erase and Replace", and "Remove Background" features provided by Runway to change the image ratio, edit image contents, and extract goal elements. Finally, imported AI-edited images into Photoshop for fine-tuning and outputting the desired final image.

Logo1

Scene Animation

Scene Animation

3D Scene Animation with Depth Map: Imported edited scene images into VoluMax7 to generate the depth map with just one click and then used the animation controller to animate 3D depth scene animation in Adobe After Effects (AE). Afterward, used Runway's "Super Slow Motion" feature to add frames to some of the scene videos to enhance the animation quality.

Logo1

Character Animation

Character Animation

CrazyTalk 8.13 was used to animate the character's lips and facial expressions. For other facial expression animation, Re:Flex allowed to wrap and morph character facial expressions by drawing on the different character facial images with AE's own splines, while Adobe Premiere (PR) was used to composite all character facial animation with body movements. Ebsynth was used to transfer the style of the character images onto the action source footage to make character action animation. Meanwhile, for some specific scenes, their green screen backgrounds were added by Runway's "Add Green Background" feature as one of the preparations for further materials integration. Adobe After Effects (AE) was used to create additional auxiliary animations such as the character's body movements.

Logo1

Audio

Audio

AI-generated background music that contains customized music length, tempo, composition, instruments, and the genre was generated by Tianyin and SOUNDRAW. Freesound was used to select copyright-free sound effects that appear in the demo. Azure's TTS function, along with its high degree of customizability for voice tone and pitch, made the narration and actor lines in the demo more in line with the story's characteristics. Adobe Audition was then used to further process all audio materials, such as reducing noise and adding reverb.

Logo1

Subtitle+ Materials Integration+ Final Export

Subtitle + Materials Integration + Final Export

The materials integration, subtitle addition, and scene conversion of the demo were all completed by PR (Adobe Premiere).

Logo1

DEMO

Clips & Music

This part shows scene clips and the AI-generated music involved in the demo and trailer. And by clicking on 'interactive panel', you can check the flexibility of our experimental pipeline by seeing corresponding content conversions.


Contact Us