Skip to content

feed-and-playback.md

Latest commit

 

History

History
119 lines (84 loc) · 3.03 KB

feed-and-playback.md

File metadata and controls

119 lines (84 loc) · 3.03 KB
Raw

Feed And Playback

The main RealtimeTTS workflow is:

stream.feed(text_or_iterator).play()

feed() queues text. play() or play_async() synthesizes and plays it.

Feed Text

Use a string when all text is already available:

stream.feed("Hello. This text is ready now.")

Use an iterator when text arrives over time:

def chunks():
    yield "The first part arrives. "
    yield "The second part follows."


stream.feed(chunks())

Each feed() call returns the stream, so chaining works:

stream.feed("Hello from RealtimeTTS.").play()

Play

play() blocks until synthesis and playback finish:

stream.feed("Speak this now.")
stream.play()

Useful parameters include:

Parameter Default Purpose
fast_sentence_fragment True Allows early synthesis before a full sentence is complete.
buffer_threshold_seconds 0.0 Waits for enough buffered audio before requesting more text.
minimum_sentence_length 10 Joins very short fragments before synthesis.
minimum_first_fragment_length 10 Minimum length for the first yielded fragment.
output_wavfile None Writes synthesized audio to a WAV file.
on_audio_chunk None Receives raw audio chunks as they are ready.
tokenizer "nltk" Sentence tokenizer for streamed text.
language "en" Sentence-splitting language.
muted False Disables local speaker playback for this call.
force_first_fragment_after_words 30 Forces the first fragment after this many words.

Play Async

play_async() starts playback in a background thread:

import time

stream.feed("Background playback.")
stream.play_async()

while stream.is_playing():
    time.sleep(0.1)

play_async() has almost the same parameters as play(), but currently defaults fast_sentence_fragment_allsentences to True; play() defaults it to False. Keep this difference in mind when comparing latency between sync and async examples.

Pause, Resume, And Stop

stream.pause()
stream.resume()
stream.stop()

Pause and resume control playback. Stop asks the active engine to stop, aborts current playback, joins the async thread if needed, and resets the internal text iterator.

Pause and resume may not behave the same with engines that use the mpv playback path for compressed audio.

Text And State

if stream.is_playing():
    print("Still speaking")

print(stream.text())

is_playing() reports whether synthesis or playback is still active. text() returns the accumulated generated text.

Inline Voice And Pause Tags

The stream can register voice-switch and pause tags. The exact voice object or identifier depends on the active engine.

stream.add_voice("narrator", voice)
stream.add_pause("short-pause", 0.25)
stream.feed("[narrator]Hello.[short-pause]Now continue.")
stream.play()

The default delimiters are [ and ]. Use set_voice_tag_delimiters() if your input format needs different markers.