Boldface Accessibility Conformance Report
WCAG Edition (Based on VPAT® Version 2.5)
Draft. Prepared from automated audits and a code review, before any hand testing with assistive technology on a real iPhone. Read "Evaluation Methods Used" before relying on a "Supports".
| Name of Product/Version | Boldface 1.0 for iPhone (iOS 26 and later) |
|---|---|
| Report Date | 17 November 2026 |
| Product Description | Boldface makes word-timed captions for videos on iPhone, on the device: Live Titles and Script mode while recording, an editor (styles, text, layout and the video's shape, cuts, described video), complete captions (speaker names, sounds and music), clips from long recordings, a review with a person's sign-off, an accessibility report, and export as a captioned video plus .srt, .vtt and .txt files. Pro is a subscription bought through the App Store. |
| Contact Information | https://boldface.clayj.app/support |
| Notes | Covers the iPhone app's own interface: first launch, Home, recording, the editor and its tools, export, the report, the review, clips, the purchase screen, Settings and My Words. Boldface 1.0 runs on iPhone only, in portrait. WCAG 2.2 is applied to a native app as the W3C's WCAG2ICT guidance describes ("page" read as "screen", "set of web pages" as "set of software programs"). The captions, descriptions and transcripts Boldface makes for people's own videos are a separate matter, covered in their own section after the tables. |
| Evaluation Methods Used | See below. |
Evaluation Methods Used
- Apple's accessibility audit, automated.
BoldfaceUITests/AccessibilityAuditTests.swiftrunsXCUIApplication.performAccessibilityAudit()with every audit type (contrast, element detection, hit region, element description, Dynamic Type, clipped text, traits) on about 50 screens and sheets, in 18 tests: the welcome and its first-launch offer; Home empty and with videos; the editor while captioning and with each of its five tools (Style, Text, Layout, Cuts, Describe); readability tips, Customize, the caption sheet, the second language sheet, edit by text (empty and with a cut); checking words, and a place speech may be missing; speakers and kept sounds, a sound's sheet, sounds being reviewed and the sound detail control; Describe (moments found, the line sheet, drafted lines, the voice sheet, the described files, recording a line, a line in your own voice); Export before and after the video is made, its files with the post kit and the package, both kits' sheets, the accessibility report (preview, complete, signed off); the review, its sign-off and the signed review; Live Titles, Script mode during and before a take, and the script sheet; picking a recording for clips and the clips found; the Pro purchase screen; Settings, My Words, adding a word and the review log. It runs at the default text size (Large) and at the largest accessibility size (Accessibility XXXL, "AX5"), on an iPhone 18 Pro Max simulator running iOS 27.0, from a Debug build; the demo scenes the App Store screenshots use stand in for the camera, for long recordings and for Apple's speech model, which the simulator doesn't have. - Every audit finding is either fixed or explained. An issue fails the test unless the test's issue handler gives a written reason for leaving it, element by element: contrast findings are re-measured from the element's own pixels with the WCAG formula; Dynamic Type and clipped-text findings on bars that deliberately stop growing (like the system's tab bar) are named, with where the full text can be read.
- What has run so far. A first run at the default size, on 16 November before this round's last fixes, audited at least 23 of the screens; its findings are the ones this report calls fixed or explained. A second run at the default size, late on 16 November with every fix in place, reached every screen listed above, the first-launch offer and both kits included (17 tests). It did not pass: it reported 198 distinct findings it couldn't excuse. 128 are Dynamic Type "partially unsupported" and 35 clipped-text predictions; the ones checked against the code so far are on text that uses Dynamic Type text styles (Form rows, sheet text) or on caption words drawn on the video at the caption style's size, but they haven't all been checked one by one. 27 are contrast findings, 10 of them with no element attached; the others are on text whose colors measure 5.4:1 or more from the design tokens but sit over glass, a translucent bar or a sheet's edge, still to be re-measured. 5 are hit-region findings on rows that aren't controls (Export's summary, the listening progress), 2 are element-detection findings with no element attached, and 1 is an element in Settings with no description; those are still to investigate. So no remark below claims a clean audit, and nothing has been audited at AX5 yet: statements about AX5 come from the code (where layouts stack, where text is capped) and from the UI tests below.
- The Format tool came after those runs. It was added on 17 November, at the top of the Layout tool. The audit's Layout screen opens the sample in its own shape, so the next run will see the Format section's Shape row, but not Fit, Fill or Split, the backgrounds, the loops or choosing a video. Those are judged here from the code and from what the Format tool's unit and UI tests check. On 17 November the audit's Format test (the Layout screen with the sample in a split screen) ran once at the default size and passed.
- So did "Speech may be missing here". Build 4 (17 November) marks places where Boldface heard talking but found no words, as a row in the Text tool's caption list. Its own test, the eighteenth, ran at the default size and reported 4 findings, all Dynamic Type "partially unsupported", on the row's text, which uses text styles (Caption, Body, Footnote) and isn't capped, as on a sound's row. Check words' count reads as shown ("1 to check"), with "1 place where speech may be missing" as its value. The row is one element for VoiceOver, with Mark as Checked among its actions and selected when Next goes to it, and its Checked button is also in the row's menu and on a swipe.
- Flash measurement. The six split-screen loops that ship with Boldface were measured frame by frame when they were made: relative luminance in 4 × 4-pixel blocks, a change of 0.1 or more where the darker side is under 0.8 counting toward a flash, and red flashes counted too. A 10° field was taken as half the loop's width (a 540 × 540 window), which is conservative on an iPhone. No window ever had 25% or more of its area flashing more than three times in a second: the most was 12.2% (Marbles), the whole frame at most 5.6% (Sorting), and no loop has a red flash. The unit tests also run the app's own check (see 2.3.1) on every loop.
- UI tests at the largest size. The UI tests for round 5's screens (the paywall, the first-launch offer, Export with the kits, Describe and the described files, the subtitle pack, clips, sounds, the review and the report) ran on an iPhone SE (3rd generation) simulator with iOS 26.5, at the default size and at AX5, keeping a screenshot of each screen: 8 of 9 passed at both sizes. The ninth, the offer, opened, but the simulator's StoreKit test products didn't load on that new simulator, so it showed its "Prices couldn't load" state (with Try again and Not now) instead of the trial; the same test passes on the iPhone 18 Pro Max simulator.
- Code review of every screen in the common tasks: accessibility labels, values, hints, traits and actions; Voice Control input labels; Reduce Motion fallbacks; Differentiate Without Color cues; text styles and every place Dynamic Type is capped.
- Grayscale. Screenshots of the audited screens converted to grayscale and reviewed for meaning carried by hue alone. (The simulator's own Color Filters setting was not used; the conversion shows the same thing, luminance only.) Done for the 26 screens of the first default-size run; nothing was carried by hue alone except the speaker stripe noted under 1.4.1. The AX5 screenshots are still to check.
- Unit tests for the logic behind the fixes: the captions default that follows Closed Captions + SDH, the labels transcript words are read with, and the step buttons beside sliders (16 tests, passing).
- Not done. No testing on a real iPhone. No hand testing with VoiceOver, Voice Control, Switch Control or Full Keyboard Access: reading order, rotors, announcements and keyboard focus are judged from the code only. Not tested at the text sizes between Large and AX5, with Bold Text, Increase Contrast, Reduce Transparency, Smart Invert or Button Shapes, in landscape (the app is portrait-only), or in any language but English (the interface is English only; captions can be in more than 20 languages).
Applicable Standards/Guidelines
| Standard/Guideline | Included In Report |
|---|---|
| Web Content Accessibility Guidelines 2.2 | Level A (Yes), Level AA (Yes), Level AAA (No) |
WCAG 2.2 includes every Level A and AA success criterion of WCAG 2.0 and 2.1 except 4.1.1 Parsing, which WCAG 2.2 removed (see the note under Table 1).
Terms
- Supports: The functionality of the product has at least one method that meets the criterion without known defects or meets with equivalent facilitation.
- Partially Supports: Some functionality of the product does not meet the criterion. Here it is also used where a criterion depends on behavior that hasn't been verified yet; the remark says which.
- Does Not Support: The majority of product functionality does not meet the criterion.
- Not Applicable: The criterion is not relevant to the product.
- Not Evaluated: The product has not been evaluated against the criterion. Used only for Level AAA.
WCAG 2.2 Report
Table 1: Success Criteria, Level A
| Criteria | Conformance Level | Remarks and Explanations |
|---|---|---|
| 1.1.1 Non-text Content | Supports | In the first audit run, the element-description check found two transcript words in edit by text read with their trailing period ("Boldface.", "phone."); they're now read as said. Every icon-only button has a label (and Voice Control input labels); decorative images are hidden from VoiceOver; the video preview reads as the current caption, the timeline as one adjustable element whose value is the caption at the playhead, style cards by their names, and the welcome's demo by a description. In the Layout tool's Format section, the shapes' little outlines and the loops' moving thumbnails are hidden from VoiceOver, and each choice is read by its name ("9 by 16, vertical", "Sunset gradient", "Marbles loop", "Your video"). Not yet checked by hand with VoiceOver. |
| 1.2.1 Audio-only and Video-only (Prerecorded) | Supports | The app's own media: the welcome's demo is silent, a still picture with caption words landing on it, and VoiceOver reads it as "A sample video, captioned on this iPhone: Every word you say turns into a caption." There is no audio-only content of the app's own. |
| 1.2.2 Captions (Prerecorded) | Supports | The app's one prerecorded video with sound, the "Meet Boldface" sample, is captioned word by word wherever it plays (Home, the editor, Export). People's own videos are theirs; see "Captions, descriptions and transcripts Boldface makes". |
| 1.2.3 Audio Description or Media Alternative (Prerecorded) | Supports | The sample's picture is an abstract animated background (rings rippling on a purple field) that carries no information beyond what is said, which is captioned. Boldface's Describe tool can make a described version of it, but the app doesn't play one by default (see Apple's Audio Descriptions label at the end). |
| 1.3.1 Info and Relationships | Supports | Headings carry the header trait (40 places, for example the welcome's headline, the Format section's Shape, Background and Other half, and each paragraph of edit by text, so VoiceOver's headings rotor moves by paragraph); lists and settings use system lists and forms; chosen options carry the selected trait; grouped rows read as one element. The audit's traits check passes. Not yet walked through by hand with VoiceOver. |
| 1.3.2 Meaningful Sequence | Supports | Reading order follows the layout; nothing is positioned out of order for VoiceOver. Judged from the code; not yet walked through by hand. |
| 1.3.3 Sensory Characteristics | Supports | Instructions name actions in words ("Tap a word to select it, then tap another…", "To delete one, swipe it or touch and hold it", VoiceOver hints such as "Swipe up or down to move between captions"), not by shape, color or position alone. |
| 1.4.1 Use of Color | Partially Supports | Statuses pair color with an icon and a word (the report's Done / Needs attention, the review's verdicts); switches show a checkmark when on; the chosen chip is filled, a change in brightness that survives grayscale, and with Differentiate Without Color it also gets a checkmark; in the Format section, the chosen shape and mode always show a checkmark, and a chosen background or loop has a checkmark badge as well as its yellow edge, with the setting on or off; sound chips are outlined with a waveform, suggestions dashed with a dot; the current caption is outlined and opens up. Gap: on the timeline, which speaker starts a caption is shown by a colored stripe; the speaker's initial replaces it only when Differentiate Without Color is on. The same caption's speaker is named in words in the Text tool and on the video. |
| 1.4.2 Audio Control | Supports | Nothing plays sound by itself: the editor, clip previews and sound review play only when asked, sound review only the sound's own moment; the welcome's demo and Live Titles' demo are silent. The Layout tool's loop thumbnails play without sound, and a split screen's other half is never heard, in the editor or in the export. |
| 2.1.1 Keyboard | Partially Supports | Not verified with a hardware keyboard or Full Keyboard Access. Controls are standard SwiftUI buttons, menus, switches and lists, and every touch-only gesture has another way in: the timeline has buttons, taps and an adjustable action; dragging captions on the video has the Layout tool's Position control and up/down actions; swipe-to-delete rows have a touch-and-hold menu or a Delete button. Whether Full Keyboard Access reaches the adjustable and custom actions (the timeline, the teleprompter) has not been checked. |
| 2.1.2 No Keyboard Trap | Supports | Every sheet and full-screen view has a visible Done, Cancel or Close button, and none captures input. Not verified with a keyboard. |
| 2.1.4 Character Key Shortcuts | Not Applicable | Boldface defines no keyboard shortcuts. |
| 2.2.1 Timing Adjustable | Supports | No time limits. Short notes (for example "All words checked") stay four seconds and are also announced to VoiceOver; nothing a person must do is timed. Live Titles' line clears after 4 seconds of silence, as live captions do; every word stays in the take. |
| 2.2.2 Pause, Stop, Hide | Supports | The welcome screen's demo caption, which cycles a word at a time beside the headline, has a Pause button at its corner (Voice Control: "Pause demo" or "Pause"); paused, it shows one caption in full and holds still. With Reduce Motion on it never cycles: it shows one caption, still, and the button isn't needed. Other moving content is a status or the task itself: the captioning progress ring, Live Titles' live words, the words landing while a video is captioned. The Pause button is new in this round and hasn't been through the audit yet. In the Layout tool's Format section, the gradient tiles and the loop tiles are still pictures; only the one chosen moves, for about 5 seconds after it's chosen or shown, and never with Reduce Motion on or with Auto-Play Video Previews off. Moving content longer than 5 seconds that starts by itself is therefore limited to the welcome's demo, which can be paused. |
| 2.3.1 Three Flashes or Below Threshold | Supports | The interface doesn't flash: the export's progress line sweeps slowly (a plain bar with Reduce Motion), and a format's gradients drift by a few percent of the frame a second. The six loops Boldface ships for split screen were measured against the general and red flash thresholds when they were made (see "Flash measurement" under Evaluation Methods), and the unit tests run each through the app's own check as well. A video a person picks for a split screen's other half is checked before it's used: every frame of the part that will play, and across its seam when it repeats, over the whole frame and over each quarter of it, against the general flash threshold (a change of a tenth of white's luminance, with the darker side under 0.8) and the red flash threshold. A video with more than three flashes in any one second is refused, and the message says why: flashing like that can cause seizures for some people. The person's own main video is theirs and isn't checked. |
| 2.4.1 Bypass Blocks | Not Applicable | Per WCAG2ICT this applies to blocks repeated across a set of software programs; Boldface is one app. Within it, VoiceOver's headings rotor moves past repeated controls. |
| 2.4.2 Page Titled | Supports | The software is titled Boldface. Screens and sheets carry titles (Settings, My Words, Customize, Edit caption, Accessibility report, Review log…); the editor is titled with the video's name. |
| 2.4.3 Focus Order | Partially Supports | Not verified by hand with VoiceOver or a keyboard. The order follows the layout in the code; sheets take focus when they open. |
| 2.4.4 Link Purpose (In Context) | Supports | Links say where they go: Privacy, Terms of Use, Contact Support, Settings › Accessibility. |
| 2.5.1 Pointer Gestures | Supports | No function needs a multipoint or path-based gesture. Swipe-to-delete rows also delete from a touch-and-hold menu (captions, sounds, described lines, recent scripts — the last added in this round) or from the item's own sheet (My Words); scrubbing the timeline also works by tapping a caption. |
| 2.5.2 Pointer Cancellation | Supports | Buttons act when the finger lifts (system behavior). Scrubbing the timeline and dragging captions follow the finger and can be dragged back before letting go. |
| 2.5.3 Label in Name | Supports | Every control's spoken name starts with its visible words, including the export screen's file chips ("Open kit", "Make package", "Open package"), corrected in this round; a ratio such as "9:16" is spoken in the words it stands for (see below). Icon-and-word controls carry Voice Control input labels with the visible words first (for example "Customize" without its PRO badge, "Hide um, uh", the tool names), and icon-only controls the words a person would say ("Settings", "Gear"; "Add word", "Add"). The Format section's shapes are spoken as words, by proportion and name ("9 by 16, vertical", "Original shape"), rather than as a ratio's digits and colon; Voice Control answers first to what's on the chip ("9:16"), then to the spoken name and a word ("Vertical"). Its tiles answer to their visible names first ("Sunset", "Marbles", "Your video"), then to "Sunset gradient" or "Photos". Not yet checked by hand with Voice Control's Show Names. |
| 2.5.4 Motion Actuation | Not Applicable | No function is operated by moving the device. |
| 3.1.1 Language of Page | Supports | The app's language is set (English); VoiceOver reads the interface in it. |
| 3.2.1 On Focus | Supports | Focusing a control never changes the screen. |
| 3.2.2 On Input | Supports | Choosing an option changes that setting only (a style restyles the captions, as its card says); nothing moves the person to another screen without a tap on a button that says so. |
| 3.2.6 Consistent Help | Not Applicable | Per WCAG2ICT this applies to a set of software programs. Within Boldface, help is in one place, Settings › Help (Contact Support). |
| 3.3.1 Error Identification | Supports | Errors are said in words, for example "Couldn't open that video" with the reason, "Prices couldn't load. Check your connection and try again.", and a video refused for a split screen because it flashes more than three times a second. |
| 3.3.2 Labels or Instructions | Supports | Every field has a visible label and a line of help under it (My Words' "Write it as" and "It might hear"; the sign-off's name and role; the script editor; the line sheet's room for words). |
| 3.3.7 Redundant Entry | Supports | Nothing asks twice. The reviewer's name is remembered for the next sign-off (and can be forgotten in Settings › Accessibility); My Words apply to every new video. |
| 4.1.2 Name, Role, Value | Supports | Standard controls supply name, role and value; custom ones set them: the timeline and the teleprompter are adjustable with their position as the value, style cards and chips carry the selected trait, switches their state, the sound detail control its level and what it adds. The Format section's shapes, modes and tiles carry the selected trait when chosen, and a Pro choice has "Pro" as its value for free users, with the hint "Free to try. Exporting with it needs Pro." The first audit run's traits check found nothing to fix. Not yet checked by hand with VoiceOver. |
Note: 4.1.1 Parsing was removed from WCAG 2.2. For WCAG 2.0 and 2.1 it applies to content in markup languages; Boldface's interface isn't markup, so it is Not Applicable there too.
Table 2: Success Criteria, Level AA
| Criteria | Conformance Level | Remarks and Explanations |
|---|---|---|
| 1.2.4 Captions (Live) | Not Applicable | The app presents no live audio of its own. Live Titles captions a person's own recording live, as they speak; that is the product's function, covered in the section after the tables. |
| 1.2.5 Audio Description (Prerecorded) | Supports | As 1.2.3: the sample's picture carries nothing that isn't said. |
| 1.3.4 Orientation | Does Not Support | Boldface 1.0 runs in portrait only, on every screen. Recording vertical video may need portrait, but Home, Settings, the editor and Export don't, so a person with an iPhone mounted in landscape can't use the app. |
| 1.3.5 Identify Input Purpose | Supports | The one field that asks about the person, the review sign-off, declares its purpose (name, job title) so it can be filled in automatically. |
| 1.4.3 Contrast (Minimum) | Supports | Text colors are 5.4:1 to 19.9:1 against their backgrounds, computed from the design tokens. The first audit run's contrast findings were on Liquid Glass bar buttons, caption chips and text under translucent bars; measured from the element's own pixels with the WCAG formula they were 6:1 to 15:1 (black on the yellow Export button 14.9:1). The clips bar's line, white 55% over a translucent bar, was moved to the brighter 66% text color to leave no doubt. Captions drawn on people's videos use the style they choose, with an outline, shadow or box. |
| 1.4.4 Resize Text | Partially Supports | Text uses Dynamic Type text styles, which grow past 300% at the largest size (AX5); the AX5 audit run is still to pass (see Evaluation Methods). Gaps: like the system's tab bar, a few bars stop growing at the largest standard size (about 135%) so the screen keeps room for the content: the editor's tool bar, the buttons in the bars under edit by text, clips, the report, the subtitle pack and both kits, Live Titles' controls and Home's header. Every one of those buttons shows in the Large Content Viewer on a touch and hold; the text beside them in those bars keeps growing, and at the accessibility sizes the bars' explanatory lines move out of them into the scrolling content. Caption chips on the timeline stop at a smaller size (the track is a fixed height) and cut short captions off; every caption is listed in full in the Text tool. On the narrowest iPhones at the default size, a few control names shrink slightly to fit beside each other; above the default size they stack instead. The Format section's Fit, Fill and Split can shrink to 80% to stay side by side up to the largest standard size, and stack at the accessibility sizes. Its "Finding your face…" chip over the video stops growing at the first accessibility size so it fits over the picture; VoiceOver reads it in full, and the Layout tool's row says the same at full size. |
| 1.4.5 Images of Text | Supports | Interface text is live text. The Boldface wordmark is a logo. Captions are drawn as text over the video in the editor; they become part of the picture only in the exported video, which is the product's output. |
| 1.4.10 Reflow | Partially Supports | In the code, at the accessibility sizes rows stack, cards go to one column and lists scroll in one direction, and nothing needs scrolling both ways except the timeline, whose layout carries meaning (time runs left to right). Gaps: not yet seen at AX5 (that run is pending); the row of style cards, each a preview of its look, scrolls sideways inside the vertically scrolling Style tool, and so do the Layout tool's rows of backgrounds and loops, each tile a preview of its picture. |
| 1.4.11 Non-text Contrast | Supports | Controls are identified by their words or symbols, which meet text contrast. State shapes that carry meaning are at least 3:1: the yellow outline and dots, switches' checkmarks, sound chips' outlines (3.8:1), and hidden filler words' dashed outline, raised in this round from 2.7:1 to 4.5:1 (7.1:1 with Increase Contrast). |
| 1.4.12 Text Spacing | Not Applicable | Per WCAG2ICT this applies where the platform lets people override text spacing; iOS has no such setting. |
| 1.4.13 Content on Hover or Focus | Not Applicable | Nothing appears on hover or focus. Menus open on a tap or a touch and hold and close with a tap outside. |
| 2.4.5 Multiple Ways | Not Applicable | Per WCAG2ICT this applies to a set of software programs. |
| 2.4.6 Headings and Labels | Supports | Headings and labels say what follows ("Sound detail", "What it is", "Words per caption", "Read aloud"). |
| 2.4.7 Focus Visible | Partially Supports | Not verified with Full Keyboard Access, which draws its own focus ring around standard controls. Custom controls (the timeline, style cards, the teleprompter) haven't been checked for a visible ring. |
| 2.4.11 Focus Not Obscured (Minimum) | Partially Supports | Not verified with a keyboard. Bars pinned at the bottom of edit by text, clips and the report sit over scrolling content, which could hide a focused row under them. |
| 2.5.7 Dragging Movements | Supports | Everything done by dragging can be done with taps: every slider has a step button at each end (added in this round, for size, position, outline, shadow and the length of a pause), captions dragged on the video can be placed with the Position control, and the timeline moves to a caption that's tapped. |
| 2.5.8 Target Size (Minimum) | Supports | Controls are at least 44 by 44 points by design. The first audit run's hit-region check found one element, the selection summary in edit by text's bar, exposed as if it could be tapped; it is now plain text. |
| 3.1.2 Language of Parts | Partially Supports | The interface is in one language. Gap: captions in a language other than English, and a video's second language line, are not marked with their language, so VoiceOver reads them with the English voice. |
| 3.2.3 Consistent Navigation | Not Applicable | Per WCAG2ICT this applies to a set of software programs. Within the app, the editor's tools, sheets' Done and Cancel, and Export stay in the same places. |
| 3.2.4 Consistent Identification | Not Applicable | Per WCAG2ICT this applies to a set of software programs. Within the app, the same function has the same name and symbol throughout. |
| 3.3.3 Error Suggestion | Supports | Where a fix is known it is said ("Check your connection and try again"; a video that can't be read says why). |
| 3.3.4 Error Prevention (Legal, Financial, Data) | Supports | Purchases go through Apple's purchase sheet, which asks for confirmation. Deleting a video, all recent scripts, or marking every word as right asks first. A review sign-off is recorded with what was signed, and a later change shows as "Changed since". |
| 3.3.8 Accessible Authentication (Minimum) | Not Applicable | Boldface has no accounts or sign-in. Purchases authenticate in Apple's own sheet. |
| 4.1.3 Status Messages | Supports | Changes that happen away from focus are announced to VoiceOver without moving focus (38 announcements in the code), for example "All words checked", "All sounds reviewed", "Boldface found 2 sounds to review", "Boldface added 2 sounds to the captions, because Closed Captions + SDH is on", "Captions moved into view", "Signed off" and "…is ready to share". Not yet heard by hand with VoiceOver. The Format section announces its results too: "Your video is ready for split screen.", a refused video with its reason, "Found your face. The crop follows it now." and "No face found, so the crop stays in the middle." |
Table 3: Success Criteria, Level AAA
Not evaluated.
Captions, descriptions and transcripts Boldface makes
This section is about what Boldface produces for people's own videos, not about the app's interface. Whether a finished video meets a standard depends on the video, on the review a person gives it, and on where it's posted; Boldface gives the tools and a record, and makes no claim about the result.
- Captions. Word-timed captions from Apple's on-device speech recognition, in the person's chosen style, burned into the exported video (open captions, which viewers can't turn off or restyle), and as .srt and .vtt files (closed captions, which a platform has to support) and a plain .txt transcript. Recognition makes mistakes, more with noise, music, accents and names. Boldface marks the words it is unsure of, and the places where it heard talking but found no words ("Speech may be missing here"), and asks a person to check them; My Words teaches it names and brands. Boldface publishes no accuracy rate.
- Complete captions. Speaker names (shown with two or more speakers; in .vtt as voice tags) and sounds and music in brackets, at three levels of detail. Sounds Boldface finds are suggestions until a person keeps them, except that when the iPhone's Closed Captions + SDH setting is on, a new video starts with the sounds it found already in its captions, and the Text tool says so; those sounds count as not yet reviewed (on the report, in the review and in the post kit's notes) until a person keeps or removes them. Existing videos keep what they have.
- Described video. Boldface finds pauses in the speech and drafts a line for each on the device; only lines a person confirms are read aloud (by a system voice or recorded in the creator's voice) and exported, as a described .mp4, an .m4a description track and a .vtt description file. Descriptions fit the pauses that exist; Boldface doesn't extend the video, so a video with few pauses may not have room to describe everything that matters.
- Shapes and split screen. A format changes the picture's shape and what fills the rest of the frame; the captions are laid out again on the new frame. When the format places them, they sit in the space under a fitted video or on a split screen's seam, and at 9:16 above the band Reels and TikTok cover. Over a moving gradient, white captions keep at least 4.5:1 at every point (the palettes are chosen for it, and a unit test checks them); the blurred video is darkened a little so captions over it read. A split screen's other half is never heard. Subtitle files and clips don't change.
- Review and report. A person can review the captions and descriptions and sign off with their name. The one-page accessibility report lists what was done and what is left, refers to the WCAG success criteria each item relates to, and states that it is not a statement of conformance.
- What the creator still decides. Checking every word and sound, choosing a style that stays readable on their video, keeping captions clear of the platform's buttons (the safe zones help), uploading the .srt or .vtt where the platform accepts it, and having the descriptions written well.
Apple's Accessibility Nutrition Labels
Apple's bar for each label is that every common task works with the feature: first launch, Home, making and editing captions, export, the purchase screen and Settings.
What the evidence supports today
- Dark Interface. Every screen is dark, including the accessibility report's preview, which was the one white screen (now dark; the exported PDF stays white, as a printed page should, and the screen says so). No bright flashes.
- Sufficient Contrast. Text 5.4:1 and up from the design tokens; the audit's contrast findings re-measured from pixels at 6:1 and up where it misreads glass. Confirm with the final audit runs at both sizes.
- Differentiate Without Color Alone. Statuses, switches, chips, sound chips and review verdicts carry a shape or word as well as color; with the setting on, speakers' initials join their colors on the timeline and chosen chips get a checkmark. The Format section's chosen shape and mode always have one.
- Captions. Boldface captions its own sample video, and new videos start with complete captions when Closed Captions + SDH is on. Each video's captions are visible on its Home card and in the editor, so people can tell what is captioned.
Check on a real iPhone first
- VoiceOver. Labels, values, actions, headings and announcements are in place, and partly audited; the common tasks haven't been done by hand with VoiceOver yet. Confirm: first launch through Get started; importing a video; checking words; moving through captions with the timeline's swipe up and down; adding a speaker and a sound; choosing a shape and a split screen's other half; Describe's lines; Export and Save to Photos; the purchase sheet; Settings.
- Voice Control. Input labels are in place; confirm with Show Names and Show Numbers that every control can be named or numbered on each common task, especially the editor's tools, the timeline and the video's controls.
- Larger Text. The fixes are in, but the audit hasn't run at the largest size yet. Run it, then confirm on a device at AX3 to AX5 and at the largest standard size that nothing overlaps, and that the Large Content Viewer shows the tool bar, Home's header and Live Titles' controls.
- Reduced Motion. Every animation has a Reduce Motion fallback in the code: the welcome's demo holds one caption still, the first-launch offer fades in instead of sliding, and in the Layout tool the preview's gradient holds still (a note says the exported video still moves) and no tile moves; with it off, only the chosen tile moves, for about 5 seconds. Confirm on a device.
- Captions, on a device: turn on Closed Captions + SDH, import a video, and confirm the sounds are kept and the Text tool says why.
Don't apply yet
- Audio Descriptions. Boldface makes described versions of people's videos, but its own sample video has no described playback that follows the system's Audio Descriptions setting, so the label doesn't apply to the app's own content yet.
Legal Disclaimer
This report is a draft based on the evaluation described above, as of the report date. It describes the accessibility of Boldface 1.0's interface and is not a legal opinion, a certification, or a statement that any video made with Boldface meets a standard.