Magnific releases Photoshop plugin with generate, upscale, retouch, and relight tools

• Magnific's Photoshop launch post puts generate, upscale, retouch, remove background, relight, reframe, feed browsing, and stock search inside Photoshop.
• The download post says the plugin is available now, and Magnific's official page lists Photoshop 2024 v25.0 or later, Windows/macOS support, a Magnific account, internet during generation, and shared web credits.
• A same-day Gemini Omni demo used Magnific to add motion graphics synced to music and subject movement while leaving the original footage untouched.
• Magnific's character-sheet thread and its Seedance 2.0 demo turned static character references into 4K consistent motion.
• The video-editor plugin thread shows the same plugin push across Premiere Pro, After Effects, DaVinci Resolve, and Final Cut Pro.

Magnific's Photoshop plugin page buries the practical requirements: Photoshop 2024 v25.0 or later, an active Magnific account, live internet during generation, and the same credits/history as the web app. Its MCP docs say Claude, ChatGPT, Cursor, Codex, VS Code, Gemini, and Hermes can call Magnific tools, while Google's Gemini API docs frame Omni Flash around multi-turn video editing.

On the official Photoshop plugin page, Magnific says the plugin runs directly on the canvas so users can generate, upscale, and retouch without exporting or switching apps.

The plugin tool list is compact:

1. Generate images and video from prompts.
2. Upscale images.
3. Remove backgrounds.
4. Reframe shots into new aspect ratios.
5. Relight finished images.
6. Browse the Magnific feed.
7. Search stock from inside Photoshop.

In the launch post, Magnific called the feed and stock layer the piece "no other plugin" offers. The useful shift is the stock-and-feed layer sitting beside generation and finishing tools, not parked in a browser tab.

According to Magnific's installation notes, the Photoshop plugin installs as a .ccx through the Creative Cloud desktop app, then appears under Plugins > Magnific.

The page lists the operating details:

• Photoshop 2024 v25.0 or later.
• Windows and macOS.
• Magnific account required.
• Internet connection required while generating.
• Results return to Photoshop.
• Generated work is saved to Magnific history.
• Plugin actions use the same credits as the web platform.

Magnific's reply about model usage said the plugin uses credits regardless of the model. The broader credits docs say image generation, video creation, photo editing, audio, and other AI actions consume credits, while stock downloads have separate limits.

The video-editor plugin thread put Magnific inside Premiere Pro, After Effects, DaVinci Resolve, and Final Cut Pro, with generation, upscaling, relighting, reframing, background removal, AI music, AI voice, and 250M stock assets.

Magnific's plugins hub now frames the product as editor-native: After Effects, Premiere Pro, Final Cut Pro, DaVinci Resolve, Photoshop, and Figma are listed on the page, with Figma marked separately on the Photoshop page's editor list.

The Magnific MCP documentation says the remote server lets an AI assistant generate images and video, train consistent characters, upscale assets, and browse generation history from chat. The same docs say everything runs on the user's Magnific account and existing credits.

Google's Gemini video docs describe Omni Flash as a fast multimodal model for video generation and conversational editing, with text, image, audio, and video inputs, character consistency, and multi-turn edits such as element replacement or perspective changes.

Magnific's examples made that concrete:

• The motion-graphics demo prompted Omni to add graphics synced to music and subject movement while keeping the footage untouched.
• The anime-style demo used stock videos as inputs, with a prompt for a controlled color palette, film texture, and subtle green undertone.
• The edit demo framed Gemini Omni as "Generate once, then change anything just by telling it what you want."
• The consistency demo claimed scene memory, plain-language edits, consistent characters and physics, and 10-second clips.
• In techhalla's UI animation workflow, Nano Banana 2 Lite created a 3D button image, Omni Flash animated it, then a reference edit swapped the play icon for a pause icon.

Magnific's Seedance 2.0 page describes the model as a 4K video generator with image, video, and audio references, camera control, multi-shot storytelling, and consistent characters.

The workflow pattern across Magnific's posts was reference first, motion second:

• The character-sheet thread used GPT 2 to create a half-elf ranger sheet, then moved into Seedance 2.0 4K.
• The 4K consistency demo said front, side, and back views matched through motion, including helmet texture.
• The storyboard demo turned a drawn action sequence into a gothic crypt video prompt with camera tracking, candlelight, film grain, and 24fps motion.
• The Fable 5 vs Sonnet 5 demo asked each model for a video prompt, then used the Magnific MCP with Seedance 2.0.
• The Fable prompt broke a 10-second teaser into five timed shots, including a stitched-fabric village, an embroidered map, a forest creature, and an end card.

Magnific's site-building thread used Library as product memory. In the Library step, Magnific said uploading a product lets the MCP, Image Generator, or any workflow know the product by name.

The landing-page prompt in the showroom step asked for an interactive product showroom using one Space for brand guidelines and four named Library products.

The video structure in the hero prompt specified two interactive videos:

1. A scroll-scrubbed hero where a centered bottle rotates 360 degrees, the cap closes, and the scene stays pinned for about three screen heights.
2. A cursor band where translucent mist gathers left or drifts right with mouse or touch movement.

The final-result post showed the finished page, and the closing post pointed creators to connect the Magnific MCP with Claude.