Each user may have up to two accepted requests in any rolling 24-hour period. This is not a reset at midnight: each request stops counting exactly 24 hours after it was submitted.

• Queued and completed requests count toward the limit.
• Failed requests do not count toward the limit.
• A rejected submission is not saved and does not consume another quota slot.
• Users with a verified int21.ai email address, and other specifically approved accounts, are exempt from this quota.

If you reach the limit, the submission message tells you how long to wait. You can continue viewing earlier results while you wait for a slot to become available.

TTS, Music, and Movie each have a separate processing queue. A new request is temporarily rejected when its selected queue has 100 or more jobs waiting or being processed.

• A busy Music queue does not prevent a TTS submission, and vice versa.
• Queue size changes as workers complete jobs, so try the submission again later.
• A capacity rejection is not saved and does not consume a quota slot.

Queue capacity protects the shared service from accepting more work than it can process in a reasonable time. It does not guarantee when an accepted request will begin or finish.

1. Open Generate, select Music, and write your Lyrics.
2. Add Instructions covering style, mood, voice, instruments, and progression.
3. Select Generate. The request appears as a card on the Generations page.
4. When it is complete, select its card to listen.

Lyrics and Instructions each have an independent 2,000-character limit.

Use a bracketed label at the beginning of each section. This gives the song a visible roadmap and makes repetition, contrast, and progression clear.

Give every section one job

• Intro: establish the atmosphere; lyrics are optional.
• Verse: introduce details, characters, place, or movement.
• Pre-Chorus: raise tension and lead naturally into the hook.
• Chorus: state the central idea in the most memorable language.
• Bridge: introduce a new perspective, image, or emotional turn.
• Final Chorus: repeat the hook, optionally with a lyrical lift.
• Outro: resolve the story or leave one final image.

Make the words singable

• Keep lines in the same section reasonably similar in length.
• Use concrete images and active verbs instead of long explanations.
• Place the strongest phrase in the chorus and repeat it consistently.
• Use punctuation to suggest natural breaths and phrasing.
• Spell unusual names or invented words phonetically when pronunciation matters.
• Avoid putting production notes inside lyric lines; place them in Instructions.

Describe a coherent musical direction rather than listing disconnected adjectives. Start broad, then add the details that should remain consistent across the song.

Genre and era + tempo and mood + emotional progression + vocal character + main instruments + rhythm and low end + production space + important section changes.

Cover the essentials

• Style: name one primary genre and, if useful, one supporting influence.
• Pace: use words such as slow, mid-tempo, driving, or state a BPM.
• Emotional arc: explain how restrained verses develop into the chorus.
• Voice: specify vocal range, tone, intensity, phrasing, and harmony.
• Arrangement: identify the lead instruments and supporting textures.
• Groove: describe drums, bass movement, swing, pulse, or rhythmic feel.
• Production: describe intimacy, width, polish, room sound, or ambience.
• Section changes: call out a stripped bridge, instrumental break, or lifted final chorus.

Prefer specific relationships

“Warm and cinematic” is a starting point. “Intimate piano-led verses gradually add strings and low percussion, opening into a wide, uplifting final chorus” explains what changes, where it changes, and why.

Avoid conflicting directions

Too many genres, moods, lead instruments, or vocal types compete for attention. Choose a primary direction and use the remaining details to support it. Describe musical qualities directly instead of asking for an imitation of a particular artist.

Lyrics

Instructions

• Every lyric section has a clear label.
• The chorus contains one memorable central phrase.
• The Instructions name a primary style and emotional direction.
• Voice, instruments, rhythm, and production character are described.
• Important changes between sections are explicit.
• Lyrics and Instructions are each within 2,000 characters.

1. Open Generate and select TTS.
2. Enter the words you want spoken, including any performance directions.
3. Select Generate. The request appears as a card on the Generations page.
4. When it is complete, select its card to listen.

The text field has a 2,000-character limit. The demo currently uses one default voice.

Write the script as you want it to sound, not as it would appear in a formal document. Read it aloud once before submitting: if a sentence is difficult to say in one breath, shorten it.

• Use short sentences and familiar words for clear delivery.
• Use commas for light pauses and full stops for stronger breaks.
• Put one idea in each sentence and use paragraph breaks between topics.
• Write abbreviations, dates, prices, and symbols the way they should be spoken.
• Rephrase unusual names or technical terms if their pronunciation is ambiguous.
• Remove headings, URLs, footnotes, and formatting that should not be read aloud.

Plain narration example

Use one of the tested cues below in square brackets and place it immediately before the words it should affect. Use the cue exactly as written. Other free-form descriptions may be ignored or produce inconsistent speech in the version currently running on this demo.

Tested cues

• Emotion: [excited]. It works best when the words themselves express excitement.
• Delivery: [whisper], [low voice], [softly], and [emphasis].
• Timing: [pause] at a deliberate break.
• Natural sounds: [laughing], [inhale], [exhale], and [gasp].

Natural-sound cues can add a noticeable and sometimes long vocalization, so use them sparingly. Keep all cues simple and purposeful, and let punctuation carry most of the pacing. Speech generation varies between runs; a cue influences the performance but does not guarantee an identical result every time.

This complete example was tested against the deployed service. The cues change volume, timing, and energy while the text remains intelligible.

• The script sounds natural when read aloud.
• Sentences are short enough to follow easily.
• Punctuation creates the intended pace.
• Names, abbreviations, numbers, and symbols will be pronounced as intended.
• Only cues listed in this guide are used, with their exact spelling.
• Cues are placed immediately before the delivery should change.
• The complete text is within 2,000 characters.