Short answer: to get singing text to speech, put your words inside a <singing>…</singing> tag and generate them with Grok text to speech. The box above already does this: Sing all text wraps everything you type, so you can paste your lines, pick one of 28 voices and press Generate. The Sing-song button loads the same line in the lighter, chanted style. Trying it needs no account.
How to make a singing voice
- Paste or type your lines in the box. The sample is already wrapped in the singing tag; replace it with your own words.
- Keep Sing all text on to sing everything. The box shows the tag around your text, so you always see what gets sent.
- Pick a voice. All 28 voices accept the singing tag; the clips above use Eve.
- Press Generate and listen. Each generation can take a different tune, so generate twice if the first one is not what you wanted. Signed in, you can download the result as an MP3.
To sing just one line in a spoken script, turn Sing all text off, select the line, open More tags under Add expression and pick <singing>. That wraps only your selection:
And then he did it again. [chuckle] <singing>Happy days are here again.</singing>
What the singing tags do
<singing> and <sing-song> are two of 26 speech tags in xAI's text to speech docs, listed under vocal style. They are wrapping tags: they change how the words between the opening and closing tag are delivered, and the tags themselves are not read out.
In our three pre-recorded clips of the same line in the same voice, the plain read is 6.5 seconds; both sung versions stretch it to 8.4 seconds and raise the median pitch by about a third. <singing> is the fuller sung delivery and <sing-song> the lilting, nursery-rhyme kind of read. There is no melody input, no backing track and no way to set notes or tempo: the model decides the tune, a cappella, from the words.
Ideas to try
These are suggestions, not tests we ran for each use:
- Jingles and taglines: a short product line inside
<singing>, followed by the details in a normal read. - Children's content:
<sing-song>for a counting rhyme or a bedtime verse, with[pause]between lines. - Characters who break into song: keep the dialogue spoken and wrap only the sung line, with a
[chuckle]or[sigh]before it. - Birthday and greeting messages: your own words inside
<singing>, then a spoken sign-off.
Tips for a better sung line
- Write lines, not paragraphs. Short phrases with a rhythm of their own sing better than long sentences.
- Use line breaks and commas where you would breathe;
[breath]and[pause]add more. - Generate more than once. The tune is not fixed, so a second take can land better than the first.
- Try two or three voices. The same lines can sound quite different from voice to voice; the voices page has every voice reading one sentence.
- Mix in other tags.
<soft>,<slow>and[giggle]work inside and around a sung phrase. The speech tags guide lists all 26; the whisper, shouting and emotion pages do the same trick with other tags.
Limits, price and privacy
Signed out, you get 3 previews a day of up to 200 characters, played on the page; a free account adds MP3 download, and a Google account raises the limit to 1,000 characters per clip. The tag characters count toward the limit. Your text goes to xAI's model through our Cloudflare function to make the audio and is not stored.
If you build your own app, the same model costs $15 per million characters through xAI's API; see the pricing calculator. There is no separate price for singing; xAI bills text to speech per character.
Last updated: