Same issue here. Root cause is in the current SDK types. SpeechAnalyzer expects an array of any SpeechModule: // SpeechAnalyzer.swift (SDK) public convenience init(modules: [any SpeechModule], options: SpeechAnalyzer.Options? = nil) SpeechTranscriber does conform (indirectly) to SpeechModule: // SpeechTranscriber.swift (SDK) @available(macOS 26.0, iOS 26.0, *) final public class SpeechTranscriber : LocaleDependentSpeechModule { public convenience init(locale: Locale, preset: SpeechTranscriber.Preset) } …but SpeechDetector is a final class with no SpeechModule conformance: // SpeechDetector.swift (SDK) @available(macOS 26.0, iOS 26.0, *) final public class SpeechDetector { public init(detectionOptions: SpeechDetector.DetectionOptions, reportResults: Bool) public convenience init() } Apple’s docs say we should be able to do: let transcriber = SpeechTranscriber(...) let speechDetector = SpeechDetector() let analyzer = SpeechAnalyzer(modules: [speechDetector, transcriber]) but currently that’s impossible, since SpeechDetector doesn’t conform to SpeechModule (and it’s final, so we can’t adapt it ourselves). Workaround: initialize SpeechAnalyzer with the transcriber only, and implement voice-activation externally (e.g., auto-stop on inactivity). Waiting on clarification/fix from Apple whether SpeechDetector is intended to conform to SpeechModule in iOS 26.