Magnific demonstrates speech-timed captions and edits in Move
Magnific shared Move prompts that time captions, punch-ins and an example-video mosaic to speech. The workflow starts with a Seedance talking clip and preserves its dialogue and performance.

TL;DR
- Move combines speech-timed captions, keyword punch-ins and transitions in Magnific’s talking-video demo.
- A nine-panel reveal uses eight supplied example clips around the speaker, following Magnific’s edit prompt.
- A separate lip-sync workflow splits low-resolution acting from HD finishing, as techhalla demonstrated with H3 Max, Magnific upscaling and Kling 4.0.
The neat trick is a laptop reveal that cues eight clips to surround the speaker. Underneath, Move uses HeyGen’s open-source HyperFrames framework to build animations as HTML and render them to MP4.
Seedance performance pass
Magnific starts with a performance before adding graphics. Its Seedance 2.5 prompt specifies a 15-second vertical video with handheld-phone framing, natural gestures and four script beats:
- 0–3 seconds: Close-up for the hook.
- 3–8 seconds: A slightly wider shot for the main point.
- 8–12 seconds: The speaker turns a laptop toward the camera, displaying abstract animated shapes without text.
- 12–15 seconds: Back to the close-up for the closing line.
The generation instructions specify natural lip sync and room tone, excluding music, captions, readable text and watermarks. Captions and graphics are reserved for Move.
Speech-timed captions and the video mosaic
Move receives the talking clip and eight example videos. Magnific’s editing instructions define the treatment:
- Preservation: Keep the speaker, original dialogue and lip sync.
- Captions: Animate them to the voice and keep them clear of the face.
- Emphasis: Add brief punch-ins on key words.
- Laptop cue: Pull back into a 3 × 3 mosaic, with the creator in the centre and the eight supplied clips around them. Preserve each clip’s proportions.
- Ending: Return to the close-up for the final line.
- Transitions: Use short zoom-blurs with soft whooshes while keeping dialogue clear.
Magnific also demonstrated the approach on a street walk-and-talk, with speech guiding the captions, zooms and transitions.
Frame comments and version branches
Uploaded assets can be referenced with @ in the prompt, according to the Academy walkthrough. Its sneaker campaign example combines a storyboard and product images with animated typography.
The demonstrated editing controls include:
- Frame comments: Pause, click a point or select an area, then describe the change. Multiple comments can be applied together.
- Version branches: Each edit preserves the previous result. Returning to an older version creates a separate branch.
- Comparison: Play two versions side by side and mark favourites.
- Brand Kit: Apply colours, fonts and a logo. An existing animation can also become a movement-style reference for the kit.
H3 Max acting, Kling finishing
Separate from the Move demonstration, creator techhalla shared an audio-driven character workflow that generates the acting at low resolution before rebuilding detail.
H3 Max at 480p compared with Magnific upscaling and Kling 4.0 at 1080p
The process described in the accompanying thread has three stages:
- H3 Max, 480p: Animate a reference image against an audio clip up to 15 seconds long. Techhalla reported generation taking about 45 seconds.
- Magnific image upscaling: Enhance the original reference image, including skin and hair detail.
- Kling 4.0, 1080p: Pair that enhanced image with the low-resolution performance clip.
Higher-resolution H3 Max output looked too plastic to techhalla, who preferred its low-resolution acting and lip sync. The detailed reference image supplies the appearance for the finishing pass.
Access and export controls
Move is available under Motion graphics in Magnific, according to its October 28 launch announcement.
- Output: The product page lists MP4 export at 1080p or 4K, up to 60 fps. The launch announcement specifies a maximum duration of 180 seconds.
- Paid generation: The Academy tutorial shows an Ask first permission mode requiring approval before generating paid assets. Auto approve permits generation without confirmation and displays costs per version.
- Render time: The tutorial estimates three to five minutes for a new motion and around two minutes for edits. Complex 3D animations and enabling motion blur can increase export time.