TGRadioBooks
Text-to-speech for poetry books: preparing lines and stanzas
Before using text-to-speech for poetry from a book, check the original line breaks, spaces between stanzas, and reading order. Then listen to a short passage in TGRadio. Preserving the structure gives the system the intended text, but does not guarantee particular pauses, poetic rhythm, or a performed interpretation.
Compare line breaks with the edition
Open the book in a reading application and choose a poem with clear line endings. Check whether the lines have collapsed into a continuous paragraph. Look especially closely at longer lines: wrapping caused by page width can resemble an intentional line break even when the author wrote a single line.
For example, the end of a line may move down in a narrow window. Widen the window or compare another view of the same edition before treating that wrap as part of the poem.
Clean up layout debris in an external editor
If you are preparing your own text file, edit it outside TGRadio. Keep the original punctuation, separate lines, and a blank line between stanzas. Remove page numbers, repeated running headers, or fragments of neighbouring text that have entered the poem. TGRadio has no built-in text or pronunciation editor.
Avoid adding a comma at every line ending to force pauses. That changes the punctuation and may change the meaning. Test the original first, and keep it separately if you make deliberate changes to a working copy.
Check the reading order of complex pages
For poetry arranged in two columns, compare the extracted text with the intended sequence on the page. A useful warning sign is a stanza that suddenly continues with a line from the neighbouring column. Notes inserted among the main lines can cause similar confusion.
A PDF with selectable text allows extraction, but does not guarantee the correct reading order. TGRadio cannot recognise text in scanned images. You will need a text version that you have the right to use.
Listen across a stanza boundary
Choose two adjoining stanzas for your test: one containing a complete sentence and another carrying a thought across a line break. Add the prepared file through Books and Upload a book, choose a voice, and press Play. Audio preparation begins after you start playback.
Listen for sentence endings and pauses that split connected phrases. Note the exact location: for example, a verb separated from its object. Visible line breaks do not necessarily produce matching pauses, so judge the result by listening.
Separate text errors from delivery preferences
If words are missing or lines are out of order, return to the source. If the words are correct but stress, pauses, or intonation do not suit you, try the same passage with another model. A different model requires a new audio version. Keep the comparison conditions unchanged.
Speech synthesis does not promise an author’s interpretation, stage performance, or exact metre. When a poem depends on unusual stress, mark that passage for another check. One awkward line is not a reason to rewrite an entire collection.