This tracker collects Google and Nvidia filings on photo and video editing, covering per-region edits, text-guided recoloring, compression, and smoother zoom previews. The batch shows camera software moving from single-pass fixes toward layered, context-aware editing built into the pipeline.
268 filings · tracking since Apr 2026 · latest Nov 2026 · updates weekly
based on all tracked filings in this watchlist · refreshes every week
Companies are racing to own the tools that let software fix, reshape, and reinvent photos and videos automatically, from erasing unwanted objects to rebuilding lighting after the shot is taken. The core fight is over which company's methods become the standard layer sitting between a camera and a finished image.
Google and Adobe carry the most weight here by a wide margin, with Google leaning into fixing flaws at capture time and Adobe concentrating on giving editors precise control over every element after the fact.
a dated entry each week this watchlist moves · older entries stay archived
Google leads this week with seven filings covering everything from glare detection to erasing text from photos. Adobe and Samsung each added two filings, touching up sharpening, zoom, and object editing.
Adobe leads this week with four filings covering color, blur repair, lip sync, and keeping AI people looking right on moving cameras. The broader push is about making AI photo and video fixes work faster, smaller, and more precisely.
Adobe leads this week with five filings, mostly around placing and lighting objects in photos. Sony, Google, Qualcomm, and Samsung each added tools to fix or reshape photos after they are taken.
Adobe and Samsung lead this week, each filing several patents around fixing and reshaping photos and videos after the fact. Google also added multiple filings, mostly around turning simple camera input into richer images and scenes.
Nvidia leads this week with three filings covering photo repair, 360-degree scenes, and video cleanup. Google and Adobe also filed, pushing the week's focus toward fixing and reshaping images before and after they are taken.
Google leads this week with four filings covering light, text, faces, and camera framing. Sony follows with three, focused on cutting out subjects, fixing harsh lighting, and keeping video edits consistent across frames.
This week's filings center on AI that selects, removes, and fills in parts of photos and video automatically. Google and Adobe are the most active, each filing three patents covering everything from text-based edits to frame rebuilding.
Adobe leads this week with seven filings covering cutting, lighting, shadows, and motion in photos and videos. Google and Adobe together push hard on making AI understand and edit specific parts of an image rather than the whole thing.
Google dominates this week with seven filings spanning video, lighting, and image repair, while Adobe adds four focused on color, reflections, and repositioning people in photos.
This week's filings show companies focused on fixing photo and video quality problems, from brightening dark spots to repairing blurry compressed video. Google and Apple each filed two patents, making them the most active this week.
Adobe leads this week with five filings covering edge cleanup, 3D scene compression, and video edit scoring. Most other companies are focused on fixing or sharpening images and video, from removing spots to rebuilding blurry photos.
This week's filings lean heavily toward AI that understands and reshapes photos and video, from changing how far away a subject looks to copying a director's camera style. Google leads the pack with three new filings, joined by Sony with two.
counts from tracked filings · focus read from each company’s own filings
20 or more filings in the last 8 weeks · 6 to 19 · under 6
the fights inside the fight · each with its three newest filings · new filings join every week
Adobe 19, Google 5, Qualcomm 4
Several companies are filing on ways to cut things out of photos and fill the empty space convincingly. The approaches range from cutting at multiple detail levels to finishing objects that were cut off at the edge of the frame.
Google 8, Adobe 4, Sony 4
These filings cover ways to correct color and lighting problems, from faces shot against bright windows to scenes lit by two different light sources at once. The approaches differ but the goal is the same: make the colors look right without the person having to fix them by hand.
Adobe 6, Nvidia 4, Sony 4
These filings tackle AI-generated or AI-processed video looking jumpy or flickery from one frame to the next. Each approach to the fix is different, but all of them target the same visual glitch.
Google 6, Samsung 3, Disney 1
These filings cover ways to train image and video AI by showing it examples rather than writing out rules. The AI learns what good output looks like by studying real photos or videos, which can make it better at tasks like sharpening, restoring, or matching a style.
Google 13, Adobe 7, Meta 1
These filings let people describe what they want in plain words and have the AI follow those instructions precisely, including editing existing photos by rewriting their text description. They cover everything from matching words to image details to keeping brand styles consistent.
Nvidia 14, Google 5, Adobe 2
These filings build a three-dimensional model of an object or scene from ordinary flat photos or video. They cover everything from single objects to full scenes with moving parts, pointing to a broad push to make 3D creation possible without special equipment.
Google 11, Samsung 5, Microsoft 2
Turning blurry, noisy, compressed or low-resolution photos into sharp ones, and colorizing old pictures.
Google 6, Samsung 5, Adobe 2
Photo tools that read the picture and offer edits before you ask: swap options, keyword taps, self-written suggestions, or a filter picked by what is in the shot.
Samsung 10, Adobe 4, Google 3
Filings that change the story of a picture: rewind or fast-forward objects, fill in missing elements, change how close a subject looks, or turn a sketch into a photo.
Adobe 7, Google 1, Meta 1
Making an added object belong in the scene: matching shadows, re-lighting real objects after the fact, dropping 3D items into photos.
Google 7, Nvidia 2, IBM 1
Picking video compression settings shot by shot, or trading realism for file size, with AI choosing the balance.
The original footage is saved so you can choose a different subject after recording ends.
A colored scribble sets both the object and its size based on how far away that spot looks.
The redrawn mouth matches the new audio while keeping how that specific person naturally moves their lips.
You can move a single object forward or backward in time without touching the rest of the photo.
every tracked filing, month by month · counts are USPTO pre-grant publications, one per publication number (an application, not a granted patent)
US 2026/0352898 A1
A learned model that scores video quality across interest, artifacts, and compression levels, automating the human review step that currently slows down distribution decisions.
US 2026/0350959 A1
Confirms the tracker's coverage of region-specific edits by showing Google's work on isolated object captioning, the system identifies and describes only user-selected areas rather than full images, extending AI control to granular spatial selection.
US 2026/0350748 A1
The race has centered on text guidance and regional control; Microsoft now targets the speed bottleneck by running AI sharpening on sparse frames and interpolating between them for real-time playback.
US 2026/0350878 A1
Google moves beyond editing individual product shots to generating full lifestyle scenes around them, eliminating the photographer and model costs that currently gate visual variety in e-commerce.
US 2026/0350769 A1
Detecting glare on document scans by analyzing overall image brightness and isolating peak highlights extends the editing toolbox beyond color and layout toward capture quality repair.
US 2026/0350752 A1
Splitting deblurring into separate passes on thumbnail and texture layers lets phones do the heavy lifting without draining the battery as fast as full-image processing would.
US 2026/0350763 A1
Within the broader race on photo editing, Google moves past the bottleneck of needing hand-paired training data by having the AI generate its own clean reference images for text removal, crucial for real-time translation overlays.
US 2026/0350745 A1
The race to automate canvas expansion, filling empty space when you reframe a photo, now includes Adobe's one-click approach, which generates missing edges after you select an output format and reposition the original image.
US 2026/0350862 A1
The race so far has focused on regional edits and color control, but Samsung's filing shifts toward fixing what zoom itself destroys: restoring texture detail that gets smoothed away during heavy cropping and magnification.
US 2026/0352945 A1
Storing the original wide frame alongside the cropped preview lets users override the phone's initial subject choice when editing, shifting focus selection from a one-time capture decision to a post-shoot adjustment.
US 2026/0352940 A1
Detecting faces at frame edges lets the camera preserve them during zoom instead of guessing where to crop, solving the common problem of losing people at the sides of group shots.
US 2026/0350944 A1
Reconstructing unseen surfaces from a single image via 3D geometry lets editors rotate objects to reveal hidden angles, extending AI photo editing beyond what the camera captured.
US 2026/0340776 A1
Users could fix multiple damage types (blur, grain, compression artifacts) in one pass instead of running separate tools. Google's approach trains a single model by having specialist networks teach it, sidestepping the scarcity of real paired training data.
US 2026/0343106 A1
Users could dial in skin tones or correct color casts in real time rather than picking from preset categories that rarely match actual lighting. This extends the tracker's coverage of per-region color control into the camera pipeline itself.
US 2026/0341007 A1
Users can remove objects by describing them in text, but the filled-in background needs to look natural. Qualcomm's filing shows how to train the AI using paired photos of the same scene to grade its reconstruction work.
US 2026/0340778 A1
Photo inpainting on-device requires shrinking models that typically live on servers. Qualcomm's approach uses a large teacher model paired with intentional noise injection to train smaller versions that can run directly on phones.
US 2026/0340981 A1
Users get preset filters that adapt to what's actually in the frame instead of applying the same color shift everywhere. Confirms the tracker's focus on context-aware edits over one-size-fits-all adjustments.
US 2026/0340794 A1
Users editing selfies could correct perspective distortion without sacrificing fine details like jewelry or skin texture, moving beyond simple face-shape corrections into optical geometry fixes.
US 2026/0341309 A1
Syncing audio edits to mouth movements in real time shifts the challenge from speech cleanup alone to maintaining visual continuity during live playback, extending the race beyond static image retouching into dynamic video correction.
US 2026/0340921 A1
Mouth sync in dubbed video usually flattens an actor's distinctive speech habits. This filing proposes redrawing lips to match new audio while preserving individual quirks like smile timing and vowel shapes.
US 2026/0341308 A1
Users could fix flubbed dialogue without re-recording, shifting video cleanup from post-production specialists to automated speech rewriting paired with lip-sync regeneration.
US 2026/0340792 A1
Users zooming into old photos would get usable detail instead of mush. Adobe's approach reconstructs from multiple reduced copies simultaneously, cutting processing time while reducing the hallucinated artifacts that single-image methods produce.
US 2026/0340355 A1
Camera motion in generated video exposes inconsistencies across viewing angles. Adobe's approach uses multiple reference photos to maintain coherent geometry when zooming or panning around a figure.
US 2026/0330703 A1
Users pasting objects into scenes can skip manual shadow and relighting work, letting the AI handle geometry rotation and lighting consistency in a single operation.
US 2026/0330616 A1
Adobe's automation layer extends beyond which tool to use: the system now routes removals based on device power and connection speed, picking faster methods on constrained hardware.
US 2026/0330702 A1
Live video redrawing to normalize distances in wide-angle conference room feeds, so remote participants see balanced framing instead of proximity-based size distortion.
US 2026/0330614 A1
Google and Nvidia have focused on guided edits and compression; Adobe pushes toward isolating subjects more precisely by analyzing photos at multiple scales before merging results into sharp boundaries.
US 2026/0330607 A1
Providing AI sharpeners with text labels and spatial maps lets the system cross-check ambiguous regions before generating detail, reducing hallucinated textures that contradict the actual subject.
US 2026/0330595 A1
Shrinking the photo first to analyze grain patterns lets the chip skip full-resolution processing and compute a custom filter per region, cutting the work needed to denoise dim shots without preset recipes.
US 2026/0332836 A1
Automatic graphic repositioning across re-recorded takes eliminates re-syncing graphics to new speaker positions and timing, extending the editing toolkit for multi-take workflows where reshoots change camera framing or delivery pace.
US 2026/0330755 A1
Material inference from multi-angle captures lets users relight shiny objects without reshooting, extending the compression and recoloring work to handle reflective surfaces and their physical properties.
US 2026/0330697 A1
Compression via selective pixel transmission extends the AI redraw toolkit beyond guided edits to bandwidth constraints, shifting the race toward inference-side reconstruction.
US 2026/0330726 A1
Samsung's filing uses AI to redraw missing body parts when you drag a cut-off person into the frame.
US 2026/0332890 A1
Sony's filing captures several versions of a shot with different focus and exposure, so the best one can be picked afterward.
US 2026/0330603 A1
Photos with closed eyes get corrected automatically by swapping in open eyes from nearby frames in the burst, eliminating the reshoot.
US 2026/0330795 A1
Adaptive frame skipping during AI upsampling lets phones smooth zoom motion without running full inference on every frame, cutting the battery hit that makes real-time zoom editing impractical today.
US 2026/0332721 A1
Sony's filing separates actors from the LED wall behind them during filming by modeling how the wall's light falls on them.
US 2026/0330765 A1
Reconstructing 3D scene geometry from flat photos to synthesize new viewpoints extends the zoom and reframing work already tracked here, pushing past crop-based solutions toward synthetic angle shifts.
US 2026/0330706 A1
Google's region-by-region word checking extends the race beyond bulk image quality into instruction-following precision, especially for multi-object scenes where earlier generators lost track of placement details.
US 2026/0323153 A1
Pinch-to-expand gestures trigger generative fill at image edges, extending the frame beyond what the camera captured rather than recovering clipped content from sensor data.
US 2026/0320850 A1
Adobe's filing swaps the food and settings in a video to match different countries, replacing reshoots and hand-gathered footage.
US 2026/0320111 A1
Google's filing generates new scenes around a single product photo, making more ad images without a new photo shoot.
US 2026/0320868 A1
Generating panoramic video from a single camera sweep requires synthesizing motion in unseen frame regions, here Google specifies how to invent plausible movement for objects that slip past the lens edges during capture.
US 2026/0320849 A1
Classifying image content to surface relevant edits narrows the tool menu from hundreds of options to a matched handful, cutting through the generic slider problem that plagues standard editors.
US 2026/0320875 A1
Embedding reconstruction data inside edited files lets users recover deleted originals without storing backup copies, a key constraint on memory-limited phones competing in the editing space.
US 2026/0320866 A1
A draggable divider lets users inspect edits in place rather than comparing separate thumbnails, with heat markers showing which regions changed most.
US 2026/0320867 A1
Google's filing applies a personal art style on the phone itself from only a few examples, with no server doing the work.
US 2026/0320865 A1
Samsung's filing lets you draw on the image to show where an object should go and how big it should be.
US 2026/0320804 A1
Dual-model approach catches tracking drift mid-sequence rather than in post-review, letting a secondary AI correct paths when the primary tracker loses lock due to occlusion or camera movement.
US 2026/0320851 A1
Google's filing removes the background and records from several sources in one step inside the camera app, instead of assembling it later.
US 2026/0320808 A1
Nvidia's filing anchors a redrawn face to the shoulders, so it stays in place instead of floating loose.
US 2026/0323139 A1
Recording multiple lens streams simultaneously lets Samsung defer zoom decisions to post-processing, where it can interpolate between angles and eliminate the visible jumps that occur when lens-swapping happens live during capture.
US 2026/0320758 A1
Among the race's speed-versus-quality tradeoff, Disney's approach splits the difference by building a fast student model trained directly by a cleaner teacher, avoiding the usual blur and ghosting that plague real-time denoise.
US 2026/0321291 A1
Samsung's filing pulls objects out of your camera roll into a sticker library you can reuse across other images.
US 2026/0321041 A1
Precise object segmentation lets users query single elements rather than whole scenes, shifting the AI's focus from broad image analysis to targeted element description.
US 2026/0311299 A1
Inferring off-camera light sources from reflective surfaces in inserted objects solves why 3D additions look pasted in, the AI reconstructs the full surrounding environment to make reflections believable without manual scene setup.
US 2026/0311300 A1
Nvidia's filing points toward localized fixes: the patent describes semantic understanding of scene content so adjustments stay confined to their target region, moving past the broad-stroke processing that degrades untouched areas.
US 2026/0311436 A1
Wide-angle capture picks up unwanted set details, so the patent applies inpainting to remove them in post-production rather than requiring physical obstruction or repositioning during filming.
US 2026/0311404 A1
Rough previews let users adjust composition and object placement before final generation, cutting the retry loop that plagues text-only prompting in competing photo tools.
US 2026/0311290 A1
Nvidia's filing rebuilds what lies outside a photo's edges, then creates a full 360-degree scene around that one shot.
US 2026/0311353 A1
Disney's filing separates see-through details like hair when turning 2D footage into 3D, where normal edge detection fails.
US 2026/0311396 A1
Synthetic face pairs let the filter isolate makeup effects from facial geometry, keeping jawlines untouched while color and texture shift. Confirms Google's push toward surgical precision in real-time beautification.
US 2026/0311295 A1
Spatial filtering to skip upscaling on uniform regions cuts processing load by roughly 75 percent, pushing AI enhancement closer to real-time mobile performance without sacrificing sharpness where it matters.
US 2026/0311380 A1
Nvidia's filing sends AI correction data along with streamed video, so blurry, compressed frames get their detail back on playback.
US 2026/0301260 A1
Google's synthetic data approach sidesteps the training bottleneck that has kept text style transfer limited to basic fonts. Automated labeling could accelerate stylized text rendering across editing tools.
US 2026/0301292 A1
Metadata tags tracking object positions across frames let the AI maintain visual consistency when applying the same edit to moving subjects, addressing the flicker problem that plagues automated video recoloring and element replacement.
US 2026/0301123 A1
Among the tracker's focus on guided edits and regional adjustments, Sony's dual-model approach adds a method for the extraction layer itself, treating subject isolation as a problem that needs multiple algorithmic views rather than one pass.
US 2026/0300715 A1
AI image generators learn by practicing how to clean up noisy pictures, but they typically treat all parts of an image the same way when adding that noise. Microsoft's new patent describes a different approach: add noise according to how each frequency band in the image actually behaves, which could make the learning process more accurate from the start.
US 2026/0303982 A1
Preprocessing frames through brightness normalization before upscaling prevents the AI model from encountering extreme lighting conditions it wasn't trained on, keeping output quality consistent across high-contrast scenes.
US 2026/0301464 A1
Extracting expression edits from paired examples rather than relying on preset filters lets the system learn face-specific changes, moving photo editing toward personalized adjustments instead of one-size-fits-all effects.
US 2026/0303951 A1
Tap-based subject detection shifts framing from center-locked zoom to user-indicated regions, automating composition around detected subjects rather than fixed frame centers.
US 2026/0301270 A1
Self-grading feedback loops let the AI compare its edits against a reference result, closing the gap between user intent and output that existing tools leave unmeasured.
US 2026/0301382 A1
Google's filing builds artificial training photos so AI can learn realistic lighting changes without matching pairs of real scenes.
US 2026/0301355 A1
Color-matching adjacent camera feeds before stitching to eliminate visible seams in 360 and surround-view captures. Shifts focus from post-stitch blending to pre-capture calibration.
US 2026/0289872 A1
Feeding a reference photo and text prompt into frame-by-frame synthesis sidesteps the manual masking work that dominates multi-frame object replacement in video editing.
US 2026/0289247 A1
Google's lightweight neural network approach fits the compression track by moving artifact removal to the decode stage rather than relying on heavier post-processing. The filing confirms the race is turning toward on-device efficiency as a core constraint.
US 2026/0289870 A1
After Google and Nvidia pushed text-guided fills, Samsung refines the upstream step: reshaping selection borders before the generator runs, aiming to eliminate the smeared edges that plague removal tools.
US 2026/0289867 A1
Automatic mode detection based on region geometry and context reduces the setup friction that currently forces users to manually select between inpainting, outpainting, and object removal workflows.
US 2026/0289754 A1
Google and Nvidia have focused on per-region edits and text-guided recoloring; Samsung shifts the problem upstream to object removal itself, using depth data to keep inpainted frames temporally consistent rather than fixing flicker after the fact.
US 2026/0289737 A1
Adobe's filing tracks where objects are across a whole clip, so edits stay consistent from frame to frame.
US 2026/0289753 A1
Dedicated artifact removal during frame prediction prevents accumulated ghosting errors in video inference, addressing quality drift that compounds across sequences.
US 2026/0292331 A1
Real-time composition scoring pushes beyond focus detection into spatial arrangement, rating how your current frame aligns with detected visual interest points and suggesting repositioning before you shoot.
US 2026/0289736 A1
Confidence-weighted blending lets the processor favor whichever exposure the neural network trusts more in each region, avoiding the flat averages that plague mixed-light shots.
US 2026/0289296 A1
Temporal transfer learning from reference objects lets the system predict aging effects on entirely different subjects, extending photo editing beyond static transformations into object lifecycle simulation.
US 2026/0289847 A1
Sparse frame sampling cuts the computational load for video regeneration by letting AI infer missing frames between keyframes rather than processing every frame in the source clip.
US 2026/0289845 A1
Parsing object descriptions and modification requests from plain text lets the system skip manual selection tools entirely, moving camera software toward conversational rather than menu-driven workflows.
US 2026/0289946 A1
Qualcomm's filing processes only selected pixels to speed up the object outlining that photo editing tools depend on.
US 2026/0279093 A1
Frame-by-frame masking of moving subjects. Adobe's skeleton-mapping approach automates the tedious work of tracking body position across shots where subjects shift and occlude.
US 2026/0278755 A1
Extends the photo recovery angle by automating vignette removal, detecting dark corner borders and synthesizing missing content rather than cropping. Fills a gap in the restoration workflow that Google and Nvidia haven't targeted yet.
US 2026/0279036 A1
Removing a standard decoder layer speeds up decompression on devices, cutting the latency between transmission and display in gallery apps and similar interfaces.
US 2026/0278805 A1
Rotoscoping automation requires training data that mixes real and synthetic footage. Disney's method generates its own labeled training set by blending computer-generated actors into real backgrounds, sidestepping the manual annotation bottleneck.
US 2026/0278754 A1
Inpainting and outpainting pull matching textures from the user's nearby shots rather than generating them, reducing artifacts when editing complex backgrounds like brick or foliage in burst sequences.
US 2026/0278879 A1
Within the race to isolate edits to specific image regions, Google's approach trains models on paired before-and-after examples to learn the precise scope of change rather than generating from scratch.
US 2026/0278875 A1
Detecting absent subjects through reference images embedded in the scene, like a photo visible in the background, extends the recoloring and object-swapping work beyond what's already captured, toward reconstructing what should have been there.
US 2026/0278751 A1
Google's filing sharpens subjects differently depending on how far away they are, instead of treating the whole frame the same.
US 2026/0278922 A1
Synthesizing novel camera angles from single images fills a gap in product photography workflows. Adobe's approach generates unseen viewpoints computationally, removing the need for reshoot or 3D modeling work.
US 2026/0278919 A1
Relighting portraits to match arbitrary lighting environments extends Adobe's work beyond basic color shifts into 3D-aware scene matching, suggesting the tracker's focus on guided edits now includes physical light simulation.
US 2026/0278907 A1
Motion inference from static scene analysis lets the system automatically determine object movement direction and speed without manual tracking, extending the photo editing race beyond recoloring and compression into video synthesis.
US 2026/0278868 A1
Within the race to simplify photo editing workflows, Adobe's unified shadow-handling system consolidates what currently requires manual detection, removal, and inpainting into a single model trained on all three operations together.
US 2026/0279310 A1
Google and Nvidia have focused on what to edit; Samsung's approach isolates where editing is needed, cutting processing load by skipping unchanged regions.
US 2026/0278804 A1
Adobe's filing cleans up the jagged outlines phone selection tools make, with a rough pass followed by a fine one.
US 2026/0281298 A1
Leveraging depth maps to automate object isolation cuts down manual masking work in mobile editing. Google's tap-to-cut approach reuses camera sensor data already present on modern phones rather than relying on computational segmentation alone.
US 2026/0268451 A1
Content-aware repair splits the workload: the model identifies faces, text, and backgrounds separately, then applies targeted fixes instead of one uniform filter across the whole frame.
US 2026/0268437 A1
Users could fix awkward poses by borrowing body positions from other photos, cutting the manual frame-by-frame work that dominates this editing category.
US 2026/0270552 A1
Automating camera preset tuning through user feedback loops shifts the editing burden from post-capture software to the camera itself, reducing manual adjustment cycles.
US 2026/0268623 A1
Repositioning subjects in photos leaves hollow voids and distorted backgrounds. Samsung's 3D model approach fills those gaps by recalculating perspective and occlusion instead of relying on flat cutouts.
US 2026/0268560 A1
Photos edited to match a specific movie or TV scene extend the race beyond color and composition into full scene synthesis. Samsung's approach automates the visual translation between your image and a target scene's lighting, costume, and setting.
US 2026/0270441 A1
Users could stream longer videos on limited data by cutting the file size that motion encoding demands. The patent confirms Google's focus on predictive models that estimate object movement before decoding, rather than measuring it after the fact.
US 2026/0270452 A1
Embedding AI corrections in the video stream itself sidesteps the need to retransmit edited footage, cutting bandwidth while letting client devices apply fixes locally as frames arrive.
US 2026/0270419 A1
Recovering detail from compressed video feeds without re-encoding opens a practical path for real-time streaming cleanup, particularly relevant for security and live-capture workflows where original files are already discarded.
US 2026/0268535 A1
Users could get stylized versions of their best moments without manually sorting through hundreds of shots or editing each one. The filing extends the race beyond region-specific edits into automated curation and generative transformation of whole libraries.
US 2026/0270401 A1
Predicting motion from a single prior frame breaks down in complex scenes. Using two reference frames instead lets the encoder capture multi-directional movement more accurately, reducing the bitrate needed for smooth playback.
US 2026/0268589 A1
Cross-location angle transfer cuts the data burden. Google's method trains on paired viewpoints in one scene, then generates unseen angles in novel environments without retraining, reducing the photo inventory needed before editing can begin.
US 2026/0268595 A1
Pre-avatar photo conditioning via text prompts lets users refine reference images before 3D conversion, sidestepping manual editing steps that typically slow avatar creation workflows.
US 2026/0268585 A1
Exposure inconsistency across multi-angle captures: Google embeds lighting correction into the 3D model during reconstruction rather than correcting each photo separately.
US 2026/0268553 A1
Most AI image generators take a single text prompt and feed the same instructions to every part of the model. Google is patenting a system that gives each layer its own separate set of instructions, potentially giving creators far finer control over the final result.
US 2026/0268453 A1
Iterative repair of degraded photos sidesteps the single-pass model's failures by cycling through guided corrections, letting each step build on the last rather than betting everything on one inference pass.
US 2026/0268559 A1
Users shooting through windows could finally get clean shots without repositioning. Adobe's dual-image approach to separating reflections from the actual scene suggests a practical path for handling one of photo editing's messiest real-world problems.
US 2026/0268706 A1
Tracking moving objects frame-to-frame without manual adjustment cuts production time on mask alignment, a core bottleneck in video post-production workflows.
US 2026/0268542 A1
Applying text prompts to redesign entire color palettes at once cuts out the manual per-element recoloring step that currently dominates design iteration workflows.
US 2026/0268452 A1
Shadow and highlight recovery in dim or backlit scenes requires data the visible spectrum alone can't capture. Sony's approach fuses infrared wavelengths recorded alongside standard photos to recover detail the camera physically missed.
US 2026/0268540 A1
Adapting color palettes to different light sources currently requires manual adjustment across every element. Adobe's filing automates that translation, letting designers shift from one lighting condition to another without rebuilding the palette by hand.
US 2026/0261674 A1
Camera hardware limits how much detail phones can store from high-contrast shots. Google's compression method routes extra color and brightness data around those constraints, keeping what older chips would normally discard.
US 2026/0260406 A1
Users get stylized photos without manual filter selection, the camera reads what's in the frame and applies treatment automatically. This extends the race beyond guided edits and recoloring into content-aware style application triggered by scene recognition.
US 2026/0260399 A1
Google and Nvidia focus on text-guided edits; Samsung's approach uses AI to isolate and enhance specific subjects without user prompts, pushing toward fully automatic subject detection.
US 2026/0260387 A1
Smoother playback on devices means recovering detail lost to compression at the decoder stage using AI rather than standard algorithms, extending Apple's push into neural reconstruction across media formats.
US 2026/0259662 A1
Within the tracker's focus on text-guided edits, this filing shifts toward direct manipulation: users gesture over regions instead of rewriting prompts, moving AI photo editing closer to traditional brush-based tools.
US 2026/0260398 A1
Color matching in generated images requires multiple constraints layered together. Microsoft's approach uses three simultaneous controls to keep AI output within brand-specific color ranges rather than guessing from text alone.
US 2026/0253183 A1
Parallel reconstruction with degradation mapping lets users adjust sharpening intensity before final output, addressing the compression and blur artifacts that plague zoom-in workflows.
US 2026/0253291 A1
Users won't need to write prompts from scratch, since the system auto-detects subjects and suggests preset edits. This fills a gap in the race where competitors still rely on typed commands.
US 2026/0254942 A1
Per-frame compression tuning shifts from fixed settings to adaptive bitrate allocation. Nvidia's approach feeds encoder decisions back into a learning loop, letting the system preserve detail during motion while cutting waste in static scenes.
US 2026/0253186 A1
Inpainting artifacts at edit boundaries require a separate refinement pass. Adobe's filing describes a dedicated cleanup model that runs after primary removal to smooth color and texture mismatches where filled areas meet original pixels.
US 2026/0253406 A1
Users could recolor marble veins separately from the base surface in one click, splitting material selections by natural pattern hierarchy instead of forcing a choice between whole-surface and manual brushing.
US 2026/0254943 A1
Video compression relies on predicting missing blocks between frames. Google's filing shows how selecting between motion-based and image-based predictions per block could improve decoder accuracy.
US 2026/0253190 A1
Measuring edit flow quality in real time. Adobe's system scores pacing and continuity across clips before human review, cutting iteration cycles in post-production work.
US 2026/0253350 A1
Filling 3D geometry gaps without manual sculpting moves photo editing into model manipulation. Text-guided replacement of object parts extends the guided-edit pattern beyond 2D surfaces into spatial structures.
US 2026/0255053 A1
Users could compose group shots in real time rather than fix framing after the fact. The patent extends the tracker's coverage of real-time guidance tools, joining Google's earlier work on spatial layout assistance during capture.
US 2026/0253184 A1
Artifact suppression during frame merge: Qualcomm embeds defect detection into the burst-capture pipeline, catching glitches before multi-shot stacking rather than correcting them after.
US 2026/0253252 A1
Users could explore video footage in 3D without manual camera tracking or scene reconstruction. Sony's approach learns lighting behavior per-point to rebuild navigable space from handheld shots.
US 2026/0253311 A1
Storing full 3D scenes demands massive memory and slows interactive editing. Adobe's compression method uses AI to encode complex spatial data into compact form, then reconstructs it on demand, potentially enabling real-time manipulation of dense 3D content.
US 2026/0253270 A1
Selective color enhancement by attention mapping shifts the race from global image treatment to region-specific tuning. The filing confirms hardware makers are moving enhancement decisions to the camera processor itself rather than post-capture software.
US 2026/0253399 A1
Users could generate video directly rather than stitching frames together, since the system refines temporal consistency across passes instead of solving each frame independently. Extends Google's diffusion work from stills into motion.
US 2026/0245334 A1
Qualcomm's filing lets you pick out an object in video by combining a description with markers you place on screen.
US 2026/0247014 A1
Splitting motion detection from image capture onto separate sensors lets the system skip sharpening static areas, which matters for video where most frames are wasted pixels anyway.
US 2026/0245167 A1
Compact neural networks embedded in the image signal processor itself let AI synthesis happen alongside sensor cleanup rather than after it, collapsing the pipeline delay that normally forces cloud offload or queuing.
US 2026/0247012 A1
Sony is patenting a way to bottle a film director's visual signature, so that new scripts can be shot, automatically, in the style of Kubrick, Kurosawa, or whoever you pick.
US 2026/0245294 A1
Google's filing gives the AI exact position data, so it can create views from angles that only a few photos show.
US 2026/0245263 A1
Breaking a photo into separate learned tokens for each object lets editors isolate and manipulate individual visual elements instead of treating the whole scene as one unit, moving the race beyond full-image operations.
US 2026/0245184 A1
Lens distortion correction in the timeline has moved from optical fixes applied at capture time to post-shot AI redrawing that lets users pick their target focal length after seeing the result, shifting the problem from hardware constraint to software choice.
US 2026/0241291 A1
Sony is patenting a way for a game to look at a photo you upload and automatically decide how an in-game object should appear, skipping the manual sliders entirely.
US 2026/0237036 A1
Users could skip manual white balance correction entirely if the software learns what neutral should look like for skin tones, clothing, or other known objects in the frame, moving beyond the pixel-averaging approach that plagues mixed-lighting scenes.
US 2026/0237124 A1
Solving flicker across frames expands the tracker's coverage from still-image edits into real-time video lighting correction, a harder constraint than per-frame processing.
US 2026/0237029 A1
Boundary frames between video chunks tend to drift in color and lighting when AI processes segments separately. Adobe's approach locks those junction points to shared parameters so adjacent chunks render with visual continuity.
US 2026/0237108 A1
Users editing longer videos gain stable scenes without forcing the AI to reload every prior frame into memory each time it renders a new one.
US 2026/0237087 A1
Extracting intrinsic color from cast shadows lets editors relight photos without recoloring the actual surface. This fills the gap between global adjustments and per-region edits by working at the optical physics layer.
US 2026/0237123 A1
Users could compose photos by sketching directly into the live preview, letting the camera generate missing elements where they draw. This shifts the editing workflow from post-production touchups to real-time scene composition.
US 2026/0237019 A1
Personalizing edits from minimal reference shots lets AI recompose subjects into new scenes without requiring massive training datasets per user, narrowing the gap between generic photo tools and custom output.
US 2026/0237206 A1
Selecting the right granularity of an object costs repeated clicks in current editors. Adobe's system generates hierarchical selections at once, letting users pick their desired scope from a single interaction without re-selecting.
US 2026/0237025 A1
Color consistency in AI-generated fill regions stays prone to subtle mismatches. Adobe's dual color-space validation during training catches errors that single-pass evaluation misses.
US 2026/0237018 A1
Users could get sharp subjects with blurred backgrounds without post-processing, since the camera chip itself would render foreground and background at mismatched resolutions and merge them before saving.
US 2026/0229000 A1
A generative model trained to infer absent objects from scene context rather than user prompts pushes the editing race toward fully automated completion, moving past text-guided tools toward systems that recognize what a scene should contain.
US 2026/0228858 A1
Qualcomm's filing separates objects while the photo is being taken, so the camera can set exposure for each one before blending.
US 2026/0230685 A1
Motion transfer between unrelated subjects extends the editing race beyond appearance changes into movement synthesis, letting users impose choreography or gesture patterns without frame-by-frame animation.
US 2026/0228949 A1
Nvidia's filing has software pick the right way to rebuild a scene before it spends computing power on it.
US 2026/0228944 A1
Google's filing marks faces as protected areas during editing, so the rest of the picture can change freely.
US 2026/0228932 A1
Nvidia's filing decides automatically where newly added objects should land in a scene, with no manual placement.
US 2026/0228856 A1
Reconstructing detail from compressed feature maps during inference sidesteps the blurriness that standard upscaling introduces when vision systems need to work with shrunk representations.
US 2026/0228508 A1
Nvidia trains generators to sharpen output by applying refinement in a secondary pass, steering models away from the statistical averaging that produces soft features in early generation stages.
US 2026/0220735 A1
Dual-level region analysis sharpens object boundary detection, a core bottleneck when users need clean selections in camera editing workflows where background separation determines final image quality.
US 2026/0220901 A1
Meta's filing handles the geometry and lighting needed to insert an object into a scene without it looking pasted in.
US 2026/0220746 A1
Splitting noise removal from detail recovery lets the pipeline restore texture in low-light shots without amplifying grain, shifting the phone's post-processing toward selective refinement rather than blanket smoothing.
US 2026/0222669 A1
Sony's filing keeps on-screen content and presenters on separate layers during recording, skipping the frame-by-frame cutout work that usually follows.
US 2026/0220741 A1
The per-region edits strand gains a key mechanism: semantic segmentation that sorts pixels by content type before applying distinct filter chains, solving the washed-out face problem.
US 2026/0220739 A1
Reconstructing detail from low-resolution input by first mapping the blur structure itself rather than guessing at missing pixels. Extends the tracker's compression and zoom preview coverage into satellite and remote imaging where pixel recovery matters most.
US 2026/0220834 A1
Auto-extracted keywords let users skip typing prompts, reducing friction in the text-guided editing workflow that other players require. Confirms the race is moving toward lower-barrier interfaces for AI recoloring and scene swaps.
US 2026/0220881 A1
Reconstructing 3D geometry from multi-angle shots lets editors reposition light sources in post rather than reshooting or manually correcting each frame, a shift from 2D per-image fixes toward object-space relighting.
US 2026/0220803 A1
Google's filing adds depth sensing so edits look physically believable across the foreground and background.
US 2026/0222693 A1
Storing a reference face model lets the system correct white balance drift by comparing live skin tones to known baseline, sidestepping the camera's general-purpose light guessing that fails under mixed lighting.
US 2026/0220898 A1
Samsung's filing ages or freshens a single object in a photo, moving it backward or forward in time without reshaping the rest of the picture.
US 2026/0220823 A1
Dual-model inference for decompression pushes beyond single-path restoration, letting competing approaches fight for artifact removal where one trained method alone leaves traces behind.
US 2026/0222692 A1
Sony's filing handles mixed lighting within a single frame, instead of picking one light source for the whole picture.
US 2026/0212661 A1
Samsung's filing trains its model by hiding tiles of an image, pushing it to understand the scene instead of matching surface patterns.
US 2026/0214134 A1
Adaptive bitrate assignment during encoding lets compression parameters shift frame-by-frame based on motion and scene content rather than applying uniform settings across an entire video, improving efficiency for mixed-pacing clips.
US 2026/0212458 A1
Google's filing finds and repairs several photo problems at once, without the user having to name each one.
US 2026/0212468 A1
Samsung's filing first asks whether a visible flaw is really a flaw, telling an intentional tilt from an accidental one before correcting it.
US 2026/0212473 A1
Samsung's filing finds the sky and faces separately and tunes contrast and color for each, so a sunset photo doesn't force a tradeoff.
US 2026/0212110 A1
Syncing transcript edits back to raw media files lets users reshape audio and video from the text layer instead of the timeline, shifting the editing surface from waveform manipulation to document-style word deletion.
US 2026/0212454 A1
Separating brightness correction from color grading during multi-shot merging lets Samsung preserve natural tones where exposures overlap, a key step toward seamless HDR and panorama output that won't need manual fixes.
US 2026/0212460 A1
Selective noise filtering based on human visual perception lets the system prioritize smooth faces and subjects over peripheral areas, reducing computational load while maintaining perceived image quality in camera processing.
US 2026/0205558 A1
Users could watch stretched or gap-filled video without the blurring and stuttering that plague current frame-insertion tools, shifting the burden from post-production timing choices to predictive motion modeling.
US 2026/0203960 A1
Users won't need to describe what they want changed; the system scans the image itself and offers preset alternatives for each detected element, speeding up the edit-select workflow that text-guided approaches require.
US 2026/0203881 A1
Merging multiple exposures into one frame requires separate AI models for highlights and shadows, a two-stage approach that avoids forcing a single model to balance conflicting exposure demands.
US 2026/0203867 A1
Poor facial visibility in dim environments blocks smooth video calls. Microsoft's filing describes real-time relighting that reconstructs underlit faces without hardware fixes or manual repositioning.
US 2026/0203996 A1
Reconstructing stable 3D scenes from video requires filtering transient objects like pedestrians and cars during model training, a step Google's method automates rather than leaving to manual cleanup afterward.
US 2026/0203977 A1
Users won't have to manually dial in edit intensity for each step, the system learns when to apply instructions heavily versus lightly across different image regions, reducing over-processing and artifact creep that text-guided edits typically suffer from.
US 2026/0203972 A1
Semantic photo reconstruction from natural language lets editors skip manual masking and regeneration workflows by describing desired changes in plain text.
US 2026/0203956 A1
Compression artifacts need selective recovery. Google's filing lets a single compressed file generate multiple detail levels on decode, pushing the recreation work to the viewing device rather than baking in one quality choice at compression time.
US 2026/0195847 A1
Adaptive training on source footage itself rather than fixed datasets lets the upscaler tune to each video's specific noise patterns and content, sharpening results beyond generic models that must work across all input types.
US 2026/0195946 A1
Predicting future lighting and color shifts within a single frame moves beyond static filters into temporal reasoning, extending the tracker's coverage beyond per-region edits into physics-aware scene dynamics.
US 2026/0192191 A1
Sony's filing focuses on keeping video and game frames consistent with each other, which matters more than the quality of any single frame.
US 2026/0197561 A1
The confidence-based white balance system extends the race beyond binary correction into graduated adjustments, letting the camera pull back when lighting conditions resist a clean answer.
US 2026/0195948 A1
Within the race's focus on per-region edits, Adobe's filing confirms the race is moving into spatial reasoning: automatically inferring light direction and shadow geometry to make composited objects blend into their new backgrounds.
US 2026/0197538 A1
IBM's filing picks and orders clips from a raw video library on command, assembling a finished video.
US 2026/0196248 A1
Swapping video backgrounds per viewer without re-rendering the foreground object adds a personalization layer to the editing race, moving beyond one-size-fits-all recoloring into dynamic content delivery that preserves the hero product across versions.
US 2026/0187769 A1
Sampling color values from adjacent frames to correct isolated color shifts in animation sequences fills a gap between Samsung's earlier text-guided recoloring work and the broader video consistency problem that Google and Nvidia have been pursuing.
US 2026/0187860 A1
Preserving 3D geometry during generation lets users reposition elements and adjust lighting after the fact, moving AI images closer to the malleable asset status of traditional 3D renders rather than locked photographs.
US 2026/0187971 A1
Separating compression noise from real image detail during streaming, Nvidia's layered detection approach identifies miscolored pixels without blurring edges, a direct fix for the banding and grid patterns that plague low-bandwidth video playback.
US 2026/0187815 A1
Frame-to-frame motion tracking lets the upscaler maintain pixel continuity across fast cuts and pans, addressing the compression artifacts that plague quick camera movements in low-res video.
US 2026/0189799 A1
Exposure bracketing during capture lets the camera recover detail across extreme lighting ratios instead of forcing a single-shot compromise, extending the editing race beyond software reconstruction to hardware capture strategy.
US 2026/0187755 A1
The compression and preview category now includes selective sharpening: Intel's approach separates intentional blur from quality loss before applying filters, letting devices sharpen only what needs it rather than degrading soft backgrounds.
US 2026/0189723 A1
Selective AI upscaling based on codec entropy signals reduces compute waste by skipping frames where compression artifacts are minimal, addressing the efficiency gap between per-frame processing and practical deployment.
US 2026/0187808 A1
Samsung's filing trains AI to restore background areas before the subject, which could give it a steadier base for rebuilding faces and details.
US 2026/0187764 A1
Microsoft's filing learns each camera sensor's own noise pattern in the dark, which could deliver cleaner raw photos before any editing.
US 2026/0187774 A1
Retrieving similar photos from a database to justify quality comparisons lets the system ground its verdicts in real examples rather than abstract metrics, moving the feedback loop closer to how humans actually evaluate camera output.
US 2026/0170616 A1
Photos with mismatched details across regions, a dog's leg disconnecting from its body, a couch edge that vanishes, would become rare if AI regions could share information during generation instead of working in isolation.
US 2026/0170722 A1
Adobe's filing lets you set exact levels for image traits before the AI generates anything, instead of fixing them with prompts afterward.
US 2026/0170654 A1
Batch object removal speeds up the core editing workflow by identifying and extracting multiple subjects simultaneously rather than forcing users through repetitive individual selections.
US 2026/0170264 A1
Automating prompt refinement shifts the bottleneck from user skill to machine interpretation, letting Adobe's software bridge the gap between casual description and generator-ready specification without manual editing cycles.
US 2026/0170323 A1
Breaking images into numbered chunks like text tokens lets Google run photo editing through the same AI models that already power its language systems, sidestepping the need for separate image-specific architectures.
US 2026/0162341 A1
Users typing casual commands like "zoom out" could get photos recomposed as if the camera had physically moved, shifting the race from editing pixels to repositioning the virtual camera itself.
US 2026/0162328 A1
Google's filing rebuilds objects that were cut off or completely blocked, then lets you move them anywhere in the frame.
US 2026/0162320 A1
Pinning down which words map to which image regions during training. Google's method uses contrastive learning to force its generator to match text descriptions with sharper spatial precision, moving beyond approximate semantic alignment.
US 2026/0154897 A1
Photo edits need to know which parts they're inventing versus observing. Google's approach runs multiple reconstructions to flag where different 3D interpretations fit the same 2D image, letting the system mark its uncertainty rather than guess confidently.
US 2026/0154857 A1
Google's filing keeps an object recognizable while applying an art style, so details don't get lost when a photo becomes a painting.
US 2026/0154872 A1
Keeping bad outputs invisible: Google embeds a quality check inside generation itself rather than showing you failures and asking you to retry.
US 2026/0156368 A1
Splitting exposure control by object instead of by frame region lets the system preserve face detail without blown-out skies, a different approach from traditional HDR's zone-based compromise.
US 2026/0154795 A1
Photos lose detail when bright areas max out the sensor. Google's method reconstructs those clipped pixels by training a generative model to invent plausible detail, rather than trying to recover what was never recorded.
US 2026/0154861 A1
Google's filing makes AI reproduce your specific subject every time, instead of a generic version of it.
US 2026/0154903 A1
Reconstructing individual objects in 3D space from limited reference images solves a bottleneck in photo-to-model workflows where generic generation fails to preserve real-world specifics like wear patterns and exact colorways.
US 2026/0154796 A1
Running the removal process twice, first to fill the gap and then to refine the boundary, Adobe targets the artifact that haunts content-aware fill: the visible seam where synthetic pixels meet the original photo.
US 2026/0148426 A1
Adobe's filing builds exclusions into image generation, so unwanted elements never appear in the first place.
US 2026/0148473 A1
Scaling 3D human generation from controlled studio data to unstructured internet photos, eliminating the need for specialized capture hardware.
US 2026/0148474 A1
Training a generative model on messy, uncontrolled 2D photos sidesteps the need for expensive 3D capture rigs, letting companies build poseable digital humans from internet-scraped images instead of controlled studio data.
US 2026/0148429 A1
Making realistic pose changes with minimal user input. Adobe's system lets editors drag sparse control points to reshape objects, cutting the manual masking and warping that currently makes physical adjustments tedious.
US 2026/0148449 A1
Splitting image generation across multiple models in sequence lets each one focus on a specific quality problem, roughness, then color, then sharpness, rather than one model juggling everything at once.
US 2026/0148427 A1
Embedding character traits into the model itself rather than regenerating them per-prompt, so the same person stays recognizable across a multi-image sequence without manual intervention between frames.
US 2026/0148475 A1
Building realistic 3D human models from random internet photos instead of studio footage would let photo editors generate new poses and angles without reshoots, a major cost cut for the AI retouching pipeline.
US 2026/0148448 A1
Google's filing, called UniTune, edits photos from plain text commands alone, with no masks, sketches or crops to prepare first.
US 2026/0148337 A1
Fitting photos to different aspect ratios without losing subject matter requires generating plausible pixels where none existed. Google's approach expands the image first, then crops, rather than choosing between distortion and loss.
US 2026/0148478 A1
Real-time radiance field rendering would let editors apply light and view changes instantly rather than waiting for re-renders. Nvidia's Morton-code sorting accelerates the voxel lookup bottleneck that currently makes these edits feel sluggish.
US 2026/0148479 A1
Nvidia's filing builds editable 3D scenes from photos in real time, which could let you change an image from several angles before flattening it back to 2D.
US 2026/0141619 A1
Reconstructing intermediate frames from multi-camera video in 3D lets editors pull usable shots from moments never actually recorded, filling gaps between the camera's capture rate and creative intent.
US 2026/0141590 A1
Users could reshape their photos by pointing to reference images instead of describing what they want, speeding up the editing workflow and reducing reliance on written prompts to guide the AI.
US 2026/0141631 A1
Gaussian splatting with learned motion tokens lets the system infer dynamic 3D geometry and movement directly from ordinary photo sequences, avoiding the depth sensors or manual annotation that currently bog down video-to-3D conversion.
US 2026/0141594 A1
Companies racing to edit photos after capture need AI that follows spatial rules. Adobe's condition-map approach shifts the control earlier, letting users blueprint where objects land before generation even starts.
US 2026/0141502 A1
Photos with flawed exposure or focus could be auto-corrected before the user ever sees them, shifting the race from editing after capture to preventing bad shots in the first place.
US 2026/0141572 A1
Keeping spatial relationships stable in generated images requires enforcing layout constraints before rendering, which Adobe's method does by mapping object positions during the generation process itself rather than correcting them after.
US 2026/0141498 A1
Generating inpainting in two stages, global layout first, then local detail refinement, lets the AI match surrounding textures and perspective without the blurry compromise of single-pass methods.
US 2026/0141593 A1
Photos edited by AI would stay visually consistent with your existing brand instead of drifting toward generic templates. Adobe's approach extracts stylistic rules from your reference images to constrain what the generator can produce.
US 2026/0141615 A1
Nvidia's filing makes a fast 3D rendering method work with the wide-angle and specialty lenses that phones and cameras actually use.
US 2026/0134545 A1
Photos could be edited faster if one AI model handles cutting out subjects, isolating skies, and detecting people in a single pass instead of running three separate networks.
US 2026/0134546 A1
Adobe's filing trains a network to clean up messy selection outlines automatically, instead of fixing the final image.
US 2026/0134260 A1
Generating clothing that physics engines can actually use cuts out the manual rigging step that currently requires technical artists to make garments move realistically in simulations.
US 2026/0134524 A1
Picking the right AI model for inpainting depends on scene complexity. Adobe's system analyzes visual difficulty first, then routes the job to whichever model handles that specific type of scene best.
US 2026/0134547 A1
A neural network that refines masks separately for each disconnected region stops Adobe's selection tool from averaging detail across a person's hand, fork, and hair as one blob.
US 2026/0134591 A1
Semantic segmentation of body parts before color assignment lets the network apply different color logic to faces versus hair versus fabric, solving the greenish-skin problem that plagues simpler colorization tools.
US 2026/0134348 A1
Photos edited after capture expand from still images into video: Nvidia's approach converts footage into riggable 3D models, letting creators animate people rather than just retouch them.
US 2026/0134602 A1
Nvidia's filing generates characters from scratch, automating the skeleton setup that normally takes separate specialist work.
US 2026/0134627 A1
Generating photorealistic edits requires 3D geometry that won't break under physics simulation, Nvidia's approach builds that constraint directly into the generation step rather than fixing problems afterward.
US 2026/0134516 A1
Within the race to move photo AI onto devices rather than cloud servers, Google is focusing on the training problem: how to pack restoration quality into models small enough for smartphones to run without lag.
US 2026/0134603 A1
Automating the rigging and skeleton step, making generated characters actually movable, pushes the race beyond static image editing into animated content creation.
US 2026/0129292 A1
Running two depth sensors in parallel lets the system reject bad focus locks when one sensor gets fooled by close subjects, improving the raw material that AI photo editing later works with.
US 2026/0127723 A1
Photos edited with this system would rely on AI trained through a structured pipeline that separates the learning process itself, potentially making the fill-in results more consistent and predictable than current erase tools.
US 2026/0129298 A1
Google's filing smooths the camera preview while you zoom, so it feels like one continuous motion instead of a jarring jump.
US 2026/0127726 A1
Filtering out photos too sharp or well-composed to need work saves processing power by skipping unnecessary enhancement cycles on already-good shots.
US 2026/0120288 A1
Applying different edits to sky, subject, and background in a single pass cuts out the manual masking step that slows down even quick phone edits.
US 2026/0120346 A1
Generative models that iterate toward a text description reveal how companies are automating color grading entirely, moving the workflow from slider adjustments to natural language prompts.
US 2026/0120334 A1
Storing a personalized AI model alongside compressed photos means the reconstruction happens on your device, not in the cloud, reducing the server load that photo editing at scale would otherwise demand.
These are patent filings, which show research direction, not confirmed features. Google has filed systems for per-region editing, text-based recoloring, compression, and zoom smoothing, but a patent does not guarantee any of these ship in a camera app. It signals where engineering time is going, not a release date.
Nvidia's filings in this batch focus on turning text descriptions or video footage into animatable 3D characters, which overlaps with photo and image work through shared AI techniques like generative modeling. It is not photo editing in the traditional sense, but the underlying methods for building and refining visual content connect the two companies' patents.
Several filings return to the same challenges: deciding which photos need heavy AI processing, editing different regions of an image with different rules, and shrinking images without losing quality on rebuild. Restoration and inpainting also appear more than once, suggesting both companies see removing flaws and filling gaps as an ongoing area worth patenting repeatedly.
Not necessarily, and not on any predictable timeline. Patents describe technical approaches companies want to protect, sometimes years before a feature reaches a device, if it ever does. This watchlist is a useful way to see what problems Google and Nvidia are actively working on, not a preview of your next software update.
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →