Poem text to speech: a voice that reads poetry with pauses
A poem lives on its pauses: the breath at the end of a line, the longer silence between stanzas, a word left hanging. That is exactly where most text-to-speech tools fail, because they read a poem like a paragraph. People want a poetry voice for a short video, a school assignment, a birthday card or simply to hear their own lines aloud.
SuperClonica has calm, warm and low voices, free to try online, and a free engine that can voice a poem line by line with real pauses. This guide shows which voices fit, how to prepare the text and how to get the pauses right.
Voices for poetry
Anna, Keeper of the Light, Ember Spirit and Muerta were recorded in Russian, so they sound most natural in Russian poetry and may keep a slight accent in other languages.
- Anna: a calm, fairly low female voice, made for letters, memories and first-person lines.
- Keeper of the Light and Ember Spirit: unhurried, with pauses, for ballads, legends and parables.
- Muerta: a warm, slightly sad low voice for farewells and love poems.
- Severus Snape: low and serious, recorded in English, for dark or dramatic verse.
- The ten stock voices of the model: clear and neutral, free for anything you publish.
Voice a stanza free online
On the page of any voice, paste a stanza of 30 to 100 characters, choose the language and press Voice it. In a few seconds you can listen and download the recording, with no account.
Prepare the poem
Read the poem aloud yourself first and mark where you breathe: those are the places that need a pause.
- Line breaks and blank lines add no pause: end every line with a comma or a full stop if it should breathe.
- An ellipsis or a dash gives only a short, comma-like pause.
- Write numbers as words and respell names the way they should sound.
- If a word gets the wrong stress, respell it or replace it with a word that keeps the rhythm.
Real pauses, line by line
For pauses that feel like poetry, voice each line on its own and put silence between them. Save the poem as poem.txt, one line per line and an empty line between stanzas, install the free engine with python -m pip install superclonica-cpu and run: import numpy as np, soundfile as sf; from superclonica import Superclonica; tts = Superclonica(); r = tts.sample_rate; gap = lambda s: np.zeros(int(r * s), np.float32); lines = [l.strip() for l in open('poem.txt', encoding='utf-8')]; sf.write('poem.wav', np.concatenate([np.concatenate([tts.synthesize(l, 'F2', 'en')[0], gap(0.6)]) if l else gap(1.2) for l in lines]), r). Every line gets 0.6 seconds of silence after it and every stanza break 1.2 seconds; change the numbers to taste. For a slower, more solemn reading, use the command-line engine with --speed 0.9.
Your poems and other people’s
Your own poems are yours to voice and publish. Classic poems whose authors died long ago are usually in the public domain, while a recent poem belongs to its author or publisher, so publishing a recording of it needs their permission. For a video you publish, use the ten stock voices, M1 to M5 and F1 to F5, free under the OpenRAIL-M license, or your own voice trained in the SuperClonica studio.