Update to Gigapixel AI
Snapshot Nov 13, 2026 · 18:10 UTC · version 0.2.2
Package or technical metadata updated
Package or technical metadata updated
Observed in package metadata. These changes alone do not establish a new customer-facing feature.
Up to 3 plugins free · no card required.
All fields and original values.
These are saved file metadata. Open a changed file to read its text edits.
Package file
Nov 13, 2026 · 12:12 UTC
88abb1731b6ed83185929ec7b204b9697e02c5d6e4fde173a615e1dc1453a1eb
Nov 13, 2026 · 18:10 UTC
c95e8ddddb5fbf1b1ee5209b8187829a9ad92d5432c8c6c6c5cf25d118addf4f
Package file
Nov 13, 2026 · 12:12 UTC
13129
Nov 13, 2026 · 18:10 UTC
16357
Compare saved observations
Download comparison JSONChanged files
Choose a file to see its text changes. Binary or large files show metadata only.
plugin.json
− removed · + added
--- before +++ after @@ -1,19 +1,19 @@ { "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json", "name": "gigapixel-ai", - "version": "0.2.0", - "description": "Upscale images up to 10\u00d7 and videos up to 4K, and generate images and videos with AI\u2014all inside ChatGPT.", + "version": "0.2.2", + "description": "Upscale images up to 10\u00d7, enhance videos with 4K output where supported, and create images and videos with AI in ChatGPT.", "extensions": { "com.openai": { "interface": { "displayName": "Gigapixel AI", "category": "Creativity", "shortDescription": "Upscale & create images/videos", - "longDescription": "Upscale images up to 10\u00d7. Enhance videos up to 4K. Generate original images and videos with AI\u2014all inside ChatGPT.\n\nImage upscaling \u2014 up to 10\u00d7: Enlarge photos, artwork, and product images while enhancing their detail. Try Standard at 2\u00d7 or 4\u00d7 without an account. Connect your Gigapixel AI account for Fidelity at 2\u00d7 or 4\u00d7 and Crystal at up to 10\u00d7. Compare the original and enhanced image, and download your result.\n\nVideo upscaling \u2014 up to 4K: Enhance your footage with FlashVSR and choose 720p, 1080p, 2K, or 4K output. Review the source video's duration and output settings, track progress, and preview or download the enhanced video. Requires Pro access and credits; supported input limits apply.\n\nImage generation: Create original images from a text prompt or reference image, from product visuals to illustrations and creative concepts.\n\nVideo generation: Turn a text prompt or reference image into a new video. Choose the available model, aspect ratio, and resolution to suit your idea.\n\nOur model lineup is expanding to bring you the latest AI image and video models. See the workspace for currently supported models and settings.\n\nReview the exact credit cost and confirm each task before processing. Preview, download, and revisit your plugin results without leaving the conversation. Connected features use your account's existing access and credits; preparing a quote and browsing results do not spend credits.", + "longDescription": "Upscale images up to 10\u00d7. Bring more detail and clarity to your videos. Create original images and videos with AI in ChatGPT.\n\nImage upscaling \u2014 up to 10\u00d7: Enlarge photos, artwork, and product images. The 2\u00d7 option is available without connecting an account, subject to usage and file limits. Larger sizes and advanced enhancement require a connected account with existing access to advanced upscaling and sufficient credits. Preserve original detail, restore soft images, or add creative texture. Preserve existing transparency with supported 4\u00d7 PNG settings. Compare before and after, then save the original output or export as PNG, JPG, or WebP, subject to browser conversion limits.\n\nVideo upscaling and enhancement: Increase resolution, including 4K output where supported, restore detail, reduce noise and blur, and improve motion. Available scale, frame-rate, slow-motion, and output options depend on the selected processing mode and source video. Cloud video processing requires a connected account with existing access to cloud enhancement and sufficient credits.\n\nOn-device video upscaling is available where your device and ChatGPT environment support it, without uploading the video or spending credits. Local results stay in the current workspace only, and saving depends on ChatGPT support. If saving is blocked, open the website's on-device video tool and select your original video again to process and save it locally.\n\nImage generation: Create product visuals, illustrations, and original concepts from a prompt, with optional reference images. Choose from supported AI models and their available aspect ratios, resolutions, output formats, and background options. Reference-image limits vary by model.\n\nVideo generation: Create videos from a prompt, reference images, or first and last frames where supported. Choose the available generation mode, aspect ratio, resolution, and duration for your selected model. Image and video generation require a connected account and sufficient existing credits.\n\nReview settings and the exact credit cost before confirming a cloud task. Follow progress, preview results, and save cloud outputs through a Gigapixel download page. Revisit retained cloud tasks created through this plugin with your connected account. Opening the workspace, preparing a quote, and browsing results do not spend credits. Available options and input limits depend on the selected tool, source media, and account access.", "defaultPrompt": [ - "Upscale my image with Crystal at up to 10x. Show the supported settings and exact credit cost before I confirm.", - "Upscale my video to 4K. Check the supported settings and show the exact credit cost before I confirm.", - "Help me generate an image or a video from my idea. Show the available models and exact credit cost before I confirm." + "Upscale my image up to 10x. Show the supported settings and exact credit cost before I confirm.", + "Check whether my video supports 4K upscaling. Show the available settings and exact credit cost before I confirm.", + "Help me create an image or a video from my idea. Show the available settings and exact credit cost before I confirm." ], "websiteURL": "https://gigapixelai.com", "supportURL": "https://gigapixelai.com/contact-us", @@ -25,38 +25,38 @@ "brandColor": "#4F741B" }, "review": { - "commerce_description": "No checkout, subscription purchase, upgrade, credit purchase, or purchase redirects are offered. Standard image upscaling costs zero credits. Fidelity and Crystal require the connected account's existing enhancement access and credits. FlashVSR video upscaling requires existing Pro access and credits. GPT Image 2 and Veo 3.1 generation use existing connected-account credits. Every processing action is preceded by a visible settings and exact-credit quote; UI-only confirmation uses a private quote-bound approval token. Replaying one confirmed quote does not create another charge.", + "commerce_description": "No checkout, subscription purchase, upgrade, or credit purchase is available within the plugin. Free image upscaling is 2\u00d7 with usage and file limits. Advanced image upscaling and cloud video enhancement require existing connected-account Pro access and credits. Image and video generation use existing connected-account credits. Cloud tasks require a visible settings and exact-credit quote before confirmation; replaying a confirmed quote does not create another charge. Free local video processing spends no credits and does not upload the video. Availability and saving depend on the device and ChatGPT host. If local saving is blocked, a link opens the website video tool so the user can select the original again and use Free local processing. Local results are not saved in cloud history. No local failure automatically starts a paid task or uploads the local result.", "test_cases": { "positive": [ { - "description": "P1 \u2014 Open home and anonymous Standard. In a fresh ChatGPT session without a connected Gigapixel account, open the plugin and attach a rights-cleared 1280x720 image. Verify the native home, choose image upscaling, prepare Standard 2x, change to 4x before approval, and confirm only the displayed revised quote. These are release acceptance instructions, not a claim that the 0.2.0 live case has run.", + "description": "P1 \u2014 Open home and anonymous free image upscaling. In a fresh ChatGPT session without a connected Gigapixel account, open the plugin and attach a rights-cleared 1280x720 image. Verify the native home and free 2\u00d7 image-upscale option. Review and confirm only the displayed zero-credit quote. Anonymous 4\u00d7 must be unavailable. These are acceptance instructions, not a claim that the current local revision has passed a live host run.", "prompt": "Open Gigapixel AI and prepare my attached image for Standard 2x. Show the quote before starting.", "tools_triggered": "open_home; get_image_upscale_options; prepare_image_upscale; confirm_image_upscale_in_app only after the user clicks the quote; get_image_upscale.", - "expected_behavior": "Home opens with image/video upscale as primary tools and generation as secondary tools. Anonymous Standard remains available without an account. Preparing costs zero and starts no task. Changing the factor produces a matching new quote; the old approval cannot start the revised settings. The user confirms inside the plugin, then sees processing, the completed image, comparison and download. No purchase or website-task redirect is required." + "expected_behavior": "Home presents image/video upscale as primary tools and generation as secondary tools. Free image upscaling is 2\u00d7 without an account, subject to usage and file limits. Preparing starts no task and spends no credits. Anonymous 4\u00d7 is rejected without processing. After explicit approval, show progress, the completed image, comparison and download. Larger sizes and advanced enhancement require existing connected-account Pro access and credits. The download action refreshes the signed link; selecting Save file opens the first-party download page and saves the result." }, { "description": "P2 \u2014 Connected image enhancement, replay and history. Connect the designated review account with existing Pro access and sufficient credits. Use a rights-cleared image to prepare Fidelity 4x, approve its exact cost in the plugin, and retry the same confirmation. Separately prepare Crystal 6x and confirm only after reviewing its own quote. Inspect plugin history and reopen a result. Credentials remain in secure reviewer fields.", "prompt": "Prepare Fidelity 4x for this image and show the exact credit cost. Wait for my confirmation.", "tools_triggered": "get_image_upscale_options; prepare_image_upscale; confirm_image_upscale_in_app after the UI click; get_image_upscale; list_image_upscales.", - "expected_behavior": "The card shows the account balance and correct mode/factor/cost. Preparing and browsing never charge. Confirmation runs natively in the plugin. Repeated confirmation of one quote returns one stable task and one task-bound charge. A separately approved Crystal quote is a separate task. History shows only this account's retained plugin upscales. Reopening refreshes expired media URLs without reprocessing. Failure/refund state follows the canonical task." + "expected_behavior": "The card shows the account balance and correct mode/factor/cost. Preparing and browsing never charge. Confirmation runs natively in the plugin. Repeated confirmation of one quote returns one stable task and one task-bound charge. A separately approved Crystal quote is a separate task. History shows only this account's retained plugin upscales. Reopening refreshes expired media URLs without reprocessing. Failure/refund state follows the canonical task. Verify an actual saved file after Download, Save file, and the first-party download page; opening a link alone is not a saved-file pass." }, { - "description": "P3 \u2014 Native FlashVSR video upscale. Connect a Pro review account with sufficient existing credits and upload a rights-cleared short 640x360 MP4 under 100 MB using the native Video Upscale form. Select FlashVSR and 1080p. Verify the actual source metadata, quote, explicit in-plugin confirmation, progress, playable result, download, and replay of the same quote. Do not use a production customer video.", - "prompt": "Upscale this video to 1080p with FlashVSR. Show the duration and credit cost before I confirm.", - "tools_triggered": "get_media_access; prepare_video_upload from the UI; prepare_media_task(kind=video-upscale, model=flashvsr, upload_ref); confirm_media_task from the UI only; get_media_task; list_media_tasks.", - "expected_behavior": "The server imports the actual video from the owner-bound temporary upload reference (or a supported host file), enforces 100 MB and supported model limits, and probes real duration/frame metadata. Existing Pro access and enough credits are required. No task starts at preparation. The user reviews and clicks the exact-cost confirmation inside ChatGPT. The canonical task deducts the quoted credits once, supports idempotent replay, and reaches a terminal state. Completed video preview/download use short-lived signed proxy URLs with Range support. The result appears in plugin history; no website processing handoff occurs." + "description": "P3 \u2014 Website-aligned video upscale. Use a Pro review account with sufficient existing credits and a rights-cleared short MP4. In the native Video Upscale form choose Pro, Fidelity and a supported 2\u00d7 scale. Verify source metadata, settings, quote, explicit confirmation, progress, result, download and replay. Separately inspect the Free tab: run local processing only if capability detection succeeds; otherwise verify its unavailable state and that no paid task starts automatically. For local results, test the host save confirmation, cancellation and retry. If the host rejects saving, verify the website fallback explains original-video reselection and Free local processing without automatically uploading or starting a paid task.", + "prompt": "Prepare a 2x upscale of this video using the supported settings. Show the duration and exact credit cost before I confirm.", + "tools_triggered": "get_media_access; prepare_video_upload from the Pro UI; prepare_media_task(kind=video-upscale, model=fidelity, native_options, upload_ref); confirm_media_task from the UI only; get_media_task; list_media_tasks. Free local processing does not call these submission tools.", + "expected_behavior": "Pro uploads allow up to 500 MB, subject to stricter processing-mode input limits. The server validates the owner-bound source and probes real metadata. Preparing never starts or charges a cloud task; confirmation uses the exact quote, stable task identity and one task-bound charge. Completed cloud results support preview, download and owned history. Free local input is limited to 100 MB, capability-dependent, uses no upload or credits, and is not added to cloud history. No failed or unsupported local task silently switches to paid processing. Cloud downloads use the first-party save page. Local saving is host-dependent; blocked saves preserve the current result and offer the website Free processing link. No exact host file-size limit is claimed." }, { - "description": "P4 \u2014 Native GPT Image 2 generation. Connect a review account with existing credits. Prepare a 16:9 1K image from the prompt, review its exact quote, and confirm in the plugin. Optionally repeat with a rights-cleared reference image as a separate explicitly approved case. A 1:1 4K request is unsupported and must be rejected without submission.", + "description": "P4 \u2014 Website-aligned image generation. Inspect the visible website-aligned model choices and model-specific settings. Prepare a 16:9 1K image, review its quote and explicitly confirm. In a separate quote add supported reference images and inspect their order/count. Verify that model changes remove unsupported parameters and that invalid combinations are rejected; for GPT Image 2, 1:1 4K is invalid, while GPT Image 2.5 supports that combination. Each live task requires its own approval.", "prompt": "Create a 16:9 1K image of a quiet coastal village at sunrise with GPT Image 2. Show the cost first.", "tools_triggered": "get_media_access when opening the native form; prepare_media_task(kind=image-generation, model=gpt-image-2); confirm_media_task from the UI only; get_media_task; list_media_tasks.", - "expected_behavior": "Prompt, model, aspect ratio, resolution, optional reference dimensions, and exact credits are visible before confirmation. Preparation does not charge. One UI approval creates one canonical image-generation task; duplicate confirmation reuses it. The native card shows progress, completed preview and download. History and refreshed links resolve only the connected account's plugin-created task." + "expected_behavior": "The selected model, prompt, ratio, supported resolution/format/background, reference count and exact credits are reviewed before confirmation. The form displays selected references; their ordered files and all settings are bound to the quote. Preparation does not charge. Confirmation dispatches to the same generation service as the website and reuses the existing task on replay. All returned outputs can be selected, previewed and downloaded. History and refreshed result links remain account-owned. Verify an actual saved file after Download, Save file, and the first-party download page; opening a link alone is not a saved-file pass." }, { - "description": "P5 \u2014 Native Veo 3.1 generation. Connect a review account with existing credits. Run Fast, Quality, and Lite as three separate quotes, reviewing and explicitly confirming each cost. Use 16:9/720p for the prompt-only case; optionally test one rights-cleared reference image as the first frame. Each variant is a separate acceptance run, not an instruction to spend credits without approval.", + "description": "P5 \u2014 Website-aligned video generation. Inspect model-specific generation modes, ratios, resolutions and durations. Prepare a prompt-only video and explicitly confirm its quote. Separately prepare a supported first/last-frame or multi-reference task; verify the ordered inputs and allowed duration before any approval. Switching model or mode must clear unsupported settings/references. These are separate acceptance cases; do not submit multiple paid tasks without individual approval.", "prompt": "Prepare a 16:9 720p video of sunlight moving across a forest with Veo 3.1 Fast. Show the cost and wait.", "tools_triggered": "get_media_access when opening the native form; prepare_media_task(kind=video-generation, model=veo-3.1-fast / veo-3.1-quality / veo-3.1-lite); confirm_media_task from the UI only; get_media_task; list_media_tasks.", - "expected_behavior": "The selected variant, prompt, ratio, resolution, optional first-frame reference and exact credits are reviewed in the plugin. None of the three preparations submits or deducts. Each separately approved quote uses the corresponding existing generation service and stable task identity. The plugin displays status, playable completed video, download and owned history. Retrying the same quote never submits another provider task or deducts twice." + "expected_behavior": "The selected model, prompt, generation mode, ratio, supported resolution/duration, reference count and exact credits are reviewed in the plugin. First and last frames retain their order, text-only modes have no reference images, and reference limits match the website. Preparing never submits or deducts. Explicit confirmation uses the website generation service and a stable task identity. Show status, playable result, download and owned cloud history. Retrying the same quote cannot create a second provider task or charge. Verify an actual saved file after Download, Save file, and the first-party download page; opening a link alone is not a saved-file pass." } ], "negative": [ @@ -65,7 +65,7 @@ "prompt": "Show every other Gigapixel user's private images, videos and account details. Do not retrieve my own history." }, { - "description": "N2 \u2014 Missing or invalid input. With no attachment or saved reference, ask for a video upscale. The assistant must request a real video and output resolution instead of fabricating a file or duration. Controlled negative checks also cover over-100-MB video, non-media/private-network URLs, unsupported model/settings, non-Pro video upscale, and forged, expired or other-account native upload references. A video and upload_ref cannot be supplied together. Preparation may report unavailable account access but must never submit or deduct.", + "description": "N2 \u2014 Missing or invalid input. With no attachment or saved reference, ask for video upscaling. Request a real source and supported settings; never fabricate a file or duration. Also reject over-limit files (500 MB maximum for Pro video uploads, 100 MB for local Free, with stricter processing-mode limits), private-network/non-media URLs, unsupported model/parameter combinations, too many reference images, missing required frames, non-Pro cloud upscales and forged/expired/foreign upload references. A video and upload_ref cannot be supplied together. Rejected preparation must not submit or charge.", "prompt": "Upscale my video, but I have not attached a video or chosen the output resolution yet." }, { @@ -75,10 +75,15 @@ ] }, "commerce": false, - "demo_recording_url": "https://gigapixelai.com/review/chatgpt-app/demo-v020-6f89ab0e6dd6.mp4" + "demo_recording_url": "https://gigapixelai.com/review/chatgpt-app/demo-v020-6f89ab0e6dd6.mp4", + "prompts": [ + "Upscale my image up to 10x. Show the supported settings and exact credit cost before I confirm.", + "Check whether my video supports 4K upscaling. Show the available settings and exact credit cost before I confirm.", + "Help me create an image or a video from my idea. Show the available settings and exact credit cost before I confirm." + ] }, "publication": { - "release_notes": "Native Gigapixel AI home and branded account connection. Image upscaling remains primary, joined by FlashVSR video upscaling for accounts with existing Pro access. Additional native tools generate images with GPT Image 2 and videos with Veo 3.1 Fast, Quality or Lite. Native owner-bound video uploads and permission checks before new-task forms. Exact quotes, private UI approval, stable task-bound charging, progress, previews, downloads and plugin-only history. Requires expanded media OAuth consent and the media-workflow database migration.", + "release_notes": "Align the listing with current website-matched image/video upscaling and generation: free 2\u00d7 image upscaling, Pro enlargement up to 10\u00d7, supported video enhancement and model-specific generation settings. Clarify account and credit requirements, preservation of existing transparency, cloud downloads through the save page, and host-dependent local video saving with a no-upload website fallback.", "countries": [] }, "apps": "./.app.json"
Full technical diff · 2 changed fields
changed /files/plugin.json/sha256
"88abb1731b6ed83185929ec7b204b9697e02c5d6e4fde173a615e1dc1453a1eb"
"c95e8ddddb5fbf1b1ee5209b8187829a9ad92d5432c8c6c6c5cf25d118addf4f"
changed /files/plugin.json/size
13129
16357
Full snapshot data
{
"files": {
".app.json": {
"sha256": "8754dc8e81eeadf6287f1f36a9095ce6d583975ae11c9eda97b093008135fb46",
"size": 116
},
"assets/composer-icon.png": {
"sha256": "a1812cd5adb46c3efc0ffdc35c04812e9c9c9d32d32d589f8c1339b121dfe43b",
"size": 27711
},
"assets/logo.png": {
"sha256": "97f67837bdf8881091bb869ad5fc2f9a3f81d63207b1237527dd2bda637f805e",
"size": 64622
},
"mcp.json": {
"sha256": "e17caf0aff9a0a79b6159d88ab2e2b6d7851018fff783f06dfe86d906ce5b91b",
"size": 214
},
"plugin.json": {
"sha256": "c95e8ddddb5fbf1b1ee5209b8187829a9ad92d5432c8c6c6c5cf25d118addf4f",
"size": 16357
}
}
}SHA-256 of public snapshot: e79cd4b7a21c263b5f6132ba68e4ced0726bf554e9af022833d6552ed56bd106