feat(tts): on-device neural voice — WAV/MP3 download + pauses - #273
Merged
Conversation
Browser (OS) voices can't be recorded, so this adds an on-device neural TTS (MMS-TTS via @huggingface/transformers, loaded through the same /hf proxy as Whisper) that generates real audio samples. Users can play the result, download WAV or MP3 (lamejs), and insert pauses with [pause] markers or blank lines (silence spliced into the buffer). OS-voice playback is unchanged. Pure tts-audio.lib.ts (WAV encode, PCM, silence, pause-split) unit-tested. New dep: @breezystack/lamejs.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds downloadable audio and pauses to Text-to-Speech, requested by a user.
Why: the browser/OS voices (Web Speech) genuinely can't be captured to a file. So this adds an on-device neural voice (MMS-TTS via
@huggingface/transformers, loaded through the same/hfproxy as Whisper) that produces real audio samples.New in the tool (OS-voice playback unchanged):
@breezystack/lamejs).[pause](or leave a blank line) and set a pause length — silence is spliced into the audio buffer.Pure
tts-audio.lib.ts(WAV encode, PCM16, silence concat, pause-split) unit-tested (7 tests). transformers + lamejs are lazy-loaded chunks (no impact on the initial page). EN + ID FAQ updated.Verify: 1365 tests green, lint 0 errors, build OK.