Alt Text Intelligence
生產力
免費 · 專為 iPad 設計。尚未針對 macOS 驗證。
Alt Text Intelligence composes alt text from photos using only your device.
The deterministic composer reads each image with the system Vision framework (labels, recognized text, animals, faces, saliency, dominant color) and writes a short, factual description. It never invents details beyond what was detected. Length presets range from a brief 150 characters to a detailed 1200, plus the platform ceilings of Bluesky, Mastodon, X, LinkedIn, and Instagram.
THREE WAYS TO USE IT
• Open Alt Text Intelligence, pick a photo, get a description.
• Share a photo from any app; the Alt Text Intelligence Share Extension drops a description into your clipboard.
• Long-press a photo in the Photos app and choose Alt Text Intelligence to describe it without leaving Photos.
Export your descriptions in the Bluesky Alt Text Dataset format (the same 279,173-row open dataset published by the author) for downstream use, citation, or fine-tuning.
A PLAYGROUND FOR THE ON-DEVICE STACK
Beyond alt text, Alt Text Intelligence ships a chat surface for the system language model with a small, growing tool catalog. The model is text-only at the prompt layer; the chat lets you watch it call tools that reach things prompts can't:
• vision_recon: reads an image you attached using the Vision framework and returns labels, OCR text, animals, face count, saliency, and dominant color.
• regex_match: applies a regular expression to a string and returns matches and capture groups.
More tools (Wikipedia summary, weather, generic web fetch) are queued for future updates. When they ship, each network call surfaces a "uses internet" indicator so you always see which tool reached out.
PRIVACY POSTURE
• The on-device language model never leaves your device.
• The deterministic alt-text composer runs fully on device.
• No accounts, no telemetry, no tracking.
• Today's build requests no network entitlements at all.
• Future network tools, when added, are scoped per call and clearly indicated.
• The Privacy Manifest declares no data collected, no tracking domains.
ACCESSIBILITY
• Atkinson Hyperlegible font option.
• Dynamic Type and VoiceOver supported across surfaces.
• W3C WAI rule citations link back to the source.
WHO IT'S FOR
The deterministic alt-text composer is for accessibility advocates, social-media posters who want WAI-compliant descriptions without sending images to a server, and developers building offline accessibility tooling. The chat-with-tools playground is for curious users and developers who want to feel the shape of the on-device CV + language model stack.
ABOUT
Alt Text Intelligence is built by Luke Steuber (Bridge City Lab LLC), speech-language pathologist and indie iOS developer. It's the iOS sibling of the Bluesky Alt Text Dataset and dr.eamer.dev/alt/.
OPEN-SOURCE NOTICES
• Atkinson Hyperlegible © 2020 Braille Institute of America, Inc., licensed under the SIL Open Font License 1.1.
• WAI / W3C accessibility guidance is paraphrased and cited.
• System Vision and on-device language model APIs are provided by Apple under the Apple Developer Program License Agreement.
NOT INCLUDED
This is not a multimodal chat client. The on-device language model is text-only at the prompt layer; image awareness comes from tools the model calls, not from passing pixels directly. Expect terse, factual alt text; the deterministic composer never invents details.
評分與評論
- 此App未收到足夠的評分或評論以顯示概況。
開發者LanguageCraft表明App的私隱慣例或包括下列資料的處理。詳情請參閲 開發者的私隱政策 。
不收集資料
開發者不會從此 App 收集任何資料。
輔助使用
開發者尚未表示此 App 支援哪些輔助使用功能。 進一步了解
資料
- 供應商
- Bridge City Lab LLC
- 大小
- 4.1 MB
- 類別
- 生產力
- 相容性
須使用 iOS 17.0 或以上版本。
- iPhone
須使用 iOS 17.0 或以上版本。 - iPad
須使用 iPadOS 17.0 或以上版本。 - Mac
須使用 macOS 14.0 或以上版本及配備 Apple M1 或以上版本晶片的 Mac。 - Apple Vision
須使用 visionOS 1.0 或以上版本。
- iPhone
- 語言
- 英文
- 年齡分級
4+
- 4+
- 版權
- © 2026 Bridge City Lab LLC