mirror of
https://github.com/anomalyco/hex.git
synced 2026-11-22 21:23:02 +01:00
* iOS keyboard: live words over the keys while dictating (stage 1, two looks)
What: while the keyboard records (in-keyboard dictation, iOS 26+), the capture's
16-bit chunks also feed Apple's SpeechAnalyzer (SpeechTranscriber with volatile and
fast results, DictationTranscriber's progressive preset for its languages), and the
words so far show over the keys: settled words in the keys' ink, the tentative tail
muted, newest line in view, whole lines at the top. Two looks behind a temporary App
Group key (debug.liveDictationLook): dim (keys at 10%) and hide. The preview takes every
touch on the keys, so nothing types while it shows; the recording bar, its cancel and
done, and the globe (drawn over the words where the keyboard has its own) work as before.
The final text still comes from the recording exactly as today; the live recogniser is
stopped when the capture stops (model lingering for the final pass).
Why: Ryan asked for the dictation to show in real time over a hidden or dimmed keyboard.
LiveTranscriptSource is one small protocol so the source can be swapped:
ParakeetLiveTranscriber (the audio so far again about once a second, a 20 s window) is
written but not wired. Never worse than today: no preview when neither switch wants it,
when less than 24 MB is left, when the language's model isn't on the iPhone (no
download), on an error, or before iOS 26. Hooks in KeyboardLocalDictation stay in the
capture's events and phase, away from the transcription call.
Verified: Studio simulator (iPhone 17 Pro iOS 27.2, SE iOS 26.5), real keyboard in a
host app with results recorded from Apple's recogniser on macOS 27.2.1 replayed (the
simulator has no speech models; LiveTranscriptReplay is simulator-only): both looks,
light and dark, upright and turned, a 60 s dictation scrolling, taps on Q and space
while recording type nothing with the preview (and "Q " without it), the globe switches
keyboards on the SE, no model -> no preview and the keys as today. Measured on macOS
with the same source: first words 0.5-1.0 s after the first word ends, word lag median
0.56-0.76 s, 2.1% of shown words rewritten, +0.4-2.6 MB in the feeding process.
code-shape.py passes. No settings UI yet (stage 2).
* iOS keyboard: Show what I say (hide look) and Type as I speak
What: look (b) only. While the keyboard records by itself the keys go and the words
so far take their place under the recording bar; the dim look and its temporary
debug.liveDictationLook key are gone. Two switches in Settings › Dictation on their
own card under the style card, iOS 26 and later only, Simplified Chinese:
- Show what I say (on by default): the preview.
- Type as I speak (off by default, "Words can change until you stop"):
LiveFieldTyping types the live words through the text document proxy, retyping only
the changed tail, and keeps exactly what it typed. It edits only while the text
before the cursor still ends with that (smart quotes allowed for; a host showing only
the end must show the part that changes); otherwise it stops and leaves the field,
and the dictation goes in at the cursor as today. When the dictation's text goes in,
KeyboardLiveTyping.handOver takes the words out first and returns the field as it was,
so spacing, Undo dictation (the whole dictation), Learn how I type and typing logs see
exactly what they did before. Cancel, a failure, nothing heard and the keyboard hiding
take the words out (a held dictation goes in at the next keyboard as before). Never
in secure, password, one-time code or card number fields, over a selection or a
two-thumb Dictate's range, or with Translate what I say on (Ryan's call).
- Never worse than today: live words only with 32 MB left once the capture runs (the
speech model's 24 MB transcription headroom plus the recogniser's share), stopped under
24 MB or on a memory warning; the preview never shows before a recogniser is running,
so a dictation without one never flashes it.
- The hand-over: the preview fades with the words it had while the keys and the
dictation's own text come in; with Parakeet now English's default the field's words
can differ slightly from the preview's (Apple's), and never flash in the preview.
Why: Ryan picked look (b) and the two switch names at the checkpoint.
Verified: Studio simulator (iPhone 17 Pro, iOS 27.2) with the replayed recogniser
results: the field typed and revised as the words came ("Quick note: Hey, I am running
about 10 minutes late. Can you grab us a table by...") and back to "Quick note:" after
Cancel; Settings › Dictation in English, dark with both on, and Simplified Chinese.
New tests in DictationLearningTests (LiveDictationTests: accumulator, joining, switch
defaults, typing and revising, stopping when the field changes or reports nothing,
private fields). README describes what the keyboard shows while recording.
code-shape.py passes.
* iOS keyboard: the preview draws only the end of a long dictation
What: LiveDictationPreviewView draws its words with one label in a clipping box, moved up by
whole lines once they overflow it, and only about the box's worth of text and three lines more
(from a word's start). It replaces the scrolling UITextView, so the words can no longer be
dragged back.
Why: on the iPhone 17 Pro a 60 s dictation with the preview held up to 27 MB more in the
keyboard (peak 103 MB, against 60 MB with the recogniser on and the preview off); the Studio
simulator showed the same, +24.5 MB.
Verified: Studio simulator, the 60 s replay: +24.5 MB (text view) -> +17.3 (capped text view)
-> +5.7 MB with the label, against +7.0 MB for the same recording without the preview. Test
for the tail in LiveDictationTests.
* iOS keyboard: the preview's tail keeps a word that starts right at the cut
Verified: LiveDictationTests on the Studio simulator.
* iOS keyboard: Type as I speak's words are replaced by the held text after the keyboard hides
What: when the keyboard hides mid-dictation, KeyboardLiveTyping notes the words it typed
(LiveDictationSettings.leftover, in the App Group) as well as taking them out. When the held
dictation goes in (inserted by itself or by Paste dictation), it replaces those words if the
text before the cursor still ends with them (LiveFieldTyping(resuming:in:)); otherwise it goes in
at the cursor as before. A new dictation clears the note.
Why: on the iPhone 17 Pro, pressing Home mid-dictation left the typed words in the field (an
app gone to the background takes no edits) and, back in the app (a new document identifier),
the strip offered Paste dictation, which would have put the dictation after them.
Verified: LiveDictationTests on the Studio simulator (9 tests).
* iOS README: what the keyboard shows while recording, measured on the iPhone 17 Pro
Adds the phone's numbers (first words about 1.2-1.3 s after speech starts; 60 s dictation 60.1 MB
with the preview against 58.1 MB without, peaks 66.5 against 60.1), the hidden-keyboard case for
Type as I speak, and what is verified where and what isn't (VoiceOver on a device, lagging hosts).
234 lines
9.5 KiB
Swift
234 lines
9.5 KiB
Swift
import AVFAudio
|
|
import CoreMedia
|
|
import Foundation
|
|
import Speech
|
|
|
|
/// Live words from Apple's speech recognition while the keyboard records: the capture's 16-bit
|
|
/// chunks go to a `SpeechAnalyzer` as they arrive, its volatile results become the tentative
|
|
/// tail and its final ones the settled words. Only for the preview: the dictation's own text
|
|
/// still comes from the recording once it stops (`KeyboardTranscription`). Its model must
|
|
/// already be on the iPhone: it never downloads one.
|
|
@available(iOS 26.0, macOS 26.0, *)
|
|
@MainActor
|
|
final class AppleLiveTranscriber: LiveTranscriptSource {
|
|
enum Module: String {
|
|
/// `SpeechTranscriber` with volatile and fast results.
|
|
case speech
|
|
/// `SpeechTranscriber` with volatile results only (the probe's comparison).
|
|
case speechSlow
|
|
/// `DictationTranscriber`'s progressive long dictation, for its languages.
|
|
case dictation
|
|
}
|
|
|
|
var onUpdate: ((LiveTranscript) -> Void)?
|
|
var onFailure: ((String) -> Void)?
|
|
/// Every result as it came (probe and log): seconds since the first chunk, final, text,
|
|
/// and the audio it covers.
|
|
var onResult: ((_ at: TimeInterval, _ isFinal: Bool, _ text: String, _ start: Double, _ end: Double) -> Void)?
|
|
|
|
private let locale: Locale
|
|
private let module: Module
|
|
private let inputFormat: AVAudioFormat
|
|
private let log: (String) -> Void
|
|
private var analyzerFormat: AVAudioFormat?
|
|
private var converter: AVAudioConverter?
|
|
private var pending: [Data] = []
|
|
private var pendingBytes = 0
|
|
private let input: AsyncStream<AnalyzerInput>
|
|
private let continuation: AsyncStream<AnalyzerInput>.Continuation
|
|
private var analyzer: SpeechAnalyzer?
|
|
private var setup: Task<Void, Never>?
|
|
private var reading: Task<Void, Never>?
|
|
private var accumulator = LiveTranscriptAccumulator()
|
|
private var stopped = false
|
|
private var firstChunkAt: Date?
|
|
/// Audio fed to the analyzer, in its own frames, for each buffer's start time.
|
|
private var fedFrames: AVAudioFramePosition = 0
|
|
/// When the analyzer was ready, from `init`, for the log.
|
|
private(set) var readyAfter: TimeInterval?
|
|
private let createdAt = Date()
|
|
|
|
/// Audio held while the analyzer gets ready; past this it gives up (`onFailure`).
|
|
static let maximumPendingSeconds: Double = 6
|
|
|
|
init?(locale: Locale, module: Module, sampleRate: Double, log: @escaping (String) -> Void) {
|
|
guard let format = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: sampleRate, channels: 1, interleaved: true) else {
|
|
return nil
|
|
}
|
|
self.locale = locale
|
|
self.module = module
|
|
self.log = log
|
|
inputFormat = format
|
|
(input, continuation) = AsyncStream.makeStream(of: AnalyzerInput.self, bufferingPolicy: .unbounded)
|
|
start()
|
|
}
|
|
|
|
private func start() {
|
|
let locale = locale
|
|
let module = module
|
|
let input = input
|
|
setup = Task { [weak self] in
|
|
let modules: [any SpeechModule]
|
|
let speech: SpeechTranscriber?
|
|
let dictation: DictationTranscriber?
|
|
switch module {
|
|
case .speech, .speechSlow:
|
|
let options: Set<SpeechTranscriber.ReportingOption> = module == .speech ? [.volatileResults, .fastResults] : [.volatileResults]
|
|
let transcriber = SpeechTranscriber(locale: locale, transcriptionOptions: [], reportingOptions: options, attributeOptions: [])
|
|
speech = transcriber
|
|
dictation = nil
|
|
modules = [transcriber]
|
|
case .dictation:
|
|
let transcriber = DictationTranscriber(locale: locale, preset: .progressiveLongDictation)
|
|
speech = nil
|
|
dictation = transcriber
|
|
modules = [transcriber]
|
|
}
|
|
guard let format = await SpeechAnalyzer.bestAvailableAudioFormat(compatibleWith: modules) else {
|
|
self?.fail("no audio format for \(locale.identifier)")
|
|
return
|
|
}
|
|
// Lingering: the dictation's own transcription, right after, finds the model loaded.
|
|
let analyzer = SpeechAnalyzer(
|
|
modules: modules,
|
|
options: SpeechAnalyzer.Options(priority: .userInitiated, modelRetention: .lingering)
|
|
)
|
|
do {
|
|
try await analyzer.prepareToAnalyze(in: format)
|
|
} catch {
|
|
self?.fail("prepare failed: \(Self.describe(error))")
|
|
return
|
|
}
|
|
guard let self, !self.stopped else {
|
|
await analyzer.cancelAndFinishNow()
|
|
return
|
|
}
|
|
self.analyzer = analyzer
|
|
self.analyzerFormat = format
|
|
if format != self.inputFormat {
|
|
self.converter = AVAudioConverter(from: self.inputFormat, to: format)
|
|
}
|
|
self.reading = Task { [weak self] in
|
|
do {
|
|
if let speech {
|
|
for try await result in speech.results {
|
|
self?.receive(String(result.text.characters), isFinal: result.isFinal, range: result.range)
|
|
}
|
|
} else if let dictation {
|
|
for try await result in dictation.results {
|
|
self?.receive(String(result.text.characters), isFinal: result.isFinal, range: result.range)
|
|
}
|
|
}
|
|
} catch {
|
|
self?.fail("results failed: \(Self.describe(error))")
|
|
}
|
|
}
|
|
do {
|
|
try await analyzer.start(inputSequence: input)
|
|
} catch {
|
|
self.fail("start failed: \(Self.describe(error))")
|
|
return
|
|
}
|
|
self.readyAfter = Date().timeIntervalSince(self.createdAt)
|
|
self.log("live: \(module.rawValue) \(locale.identifier) ready in \(Int((self.readyAfter ?? 0) * 1000)) ms, \(Int(format.sampleRate)) Hz")
|
|
let held = self.pending
|
|
self.pending = []
|
|
self.pendingBytes = 0
|
|
held.forEach { self.feed($0) }
|
|
}
|
|
}
|
|
|
|
func append(_ pcm: Data) {
|
|
guard !stopped, !pcm.isEmpty else { return }
|
|
if firstChunkAt == nil { firstChunkAt = Date() }
|
|
guard analyzerFormat != nil else {
|
|
pending.append(pcm)
|
|
pendingBytes += pcm.count
|
|
if Double(pendingBytes / 2) / inputFormat.sampleRate > Self.maximumPendingSeconds {
|
|
fail("not ready after \(Int(Self.maximumPendingSeconds)) s of audio")
|
|
}
|
|
return
|
|
}
|
|
feed(pcm)
|
|
}
|
|
|
|
func stop() {
|
|
guard !stopped else { return }
|
|
stopped = true
|
|
onUpdate = nil
|
|
continuation.finish()
|
|
setup?.cancel()
|
|
reading?.cancel()
|
|
pending = []
|
|
if let analyzer {
|
|
Task { await analyzer.cancelAndFinishNow() }
|
|
}
|
|
analyzer = nil
|
|
converter = nil
|
|
}
|
|
|
|
private func feed(_ pcm: Data) {
|
|
let frames = pcm.count / 2
|
|
guard frames > 0, let format = analyzerFormat,
|
|
let source = AVAudioPCMBuffer(pcmFormat: inputFormat, frameCapacity: AVAudioFrameCount(frames)) else { return }
|
|
source.frameLength = AVAudioFrameCount(frames)
|
|
pcm.withUnsafeBytes { raw in
|
|
guard let base = raw.baseAddress, let target = source.int16ChannelData?[0] else { return }
|
|
memcpy(target, base, frames * 2)
|
|
}
|
|
let buffer: AVAudioPCMBuffer
|
|
if let converter {
|
|
let capacity = AVAudioFrameCount((Double(frames) * format.sampleRate / inputFormat.sampleRate).rounded(.up) + 32)
|
|
guard let converted = AVAudioPCMBuffer(pcmFormat: format, frameCapacity: capacity) else { return }
|
|
let once = OneShotInput(source)
|
|
var error: NSError?
|
|
converter.convert(to: converted, error: &error) { _, status in once.next(status) }
|
|
guard error == nil, converted.frameLength > 0 else { return }
|
|
buffer = converted
|
|
} else {
|
|
buffer = source
|
|
}
|
|
let start = CMTime(value: fedFrames, timescale: CMTimeScale(format.sampleRate))
|
|
fedFrames += AVAudioFramePosition(buffer.frameLength)
|
|
continuation.yield(AnalyzerInput(buffer: buffer, bufferStartTime: start))
|
|
}
|
|
|
|
private func receive(_ text: String, isFinal: Bool, range: CMTimeRange) {
|
|
guard !stopped else { return }
|
|
let at = firstChunkAt.map { Date().timeIntervalSince($0) } ?? 0
|
|
onResult?(at, isFinal, text, range.start.seconds, range.end.seconds)
|
|
onUpdate?(accumulator.apply(text, isFinal: isFinal))
|
|
}
|
|
|
|
private func fail(_ reason: String) {
|
|
guard !stopped else { return }
|
|
log("live: \(module.rawValue) gave up: \(reason)")
|
|
let failure = onFailure
|
|
stop()
|
|
failure?(reason)
|
|
}
|
|
|
|
nonisolated private static func describe(_ error: Error) -> String {
|
|
let ns = error as NSError
|
|
return "\(ns.domain) \(ns.code) \(ns.localizedDescription)"
|
|
}
|
|
}
|
|
|
|
/// Hands `AVAudioConverter` one buffer, then says there's no more for now.
|
|
final class OneShotInput: @unchecked Sendable {
|
|
private var buffer: AVAudioPCMBuffer?
|
|
|
|
init(_ buffer: AVAudioPCMBuffer) {
|
|
self.buffer = buffer
|
|
}
|
|
|
|
func next(_ status: UnsafeMutablePointer<AVAudioConverterInputStatus>) -> AVAudioBuffer? {
|
|
guard let buffer else {
|
|
status.pointee = .noDataNow
|
|
return nil
|
|
}
|
|
self.buffer = nil
|
|
status.pointee = .haveData
|
|
return buffer
|
|
}
|
|
}
|