Every tool I tried wanted the upload first. SpeakText does the whole job on the phone: MP4, M4A, MP3 or WAV in, word-level-timed SRT subtitles or plain TXT out. Voice activity detection, punctuation restoration and inference all run locally on sherpa-onnx. 35 languages, no cloud API, works in airplane mode. It keeps running with the screen locked and resumes if Android kills it. Editing a subtitle line seeks the player to that moment. 3-day free trial.