Phase TTS

Phase TTS is designed for long narration, audiobooks, and scripts that need paragraph-by-paragraph handling.

Suitable Text

Text Type	Fit
Narration manuscript	Good fit; split by blank lines or paragraph boundaries.
Novel chapter with a small amount of dialogue	Good fit; split first, then assign characters manually.
Heavy multi-character dialogue	Prefer Dialogue TTS and clean speaker labels before import.
Very long book	Split by chapter and process one file or project at a time.

Basic Flow

1. Create a project.
2. Select a voice bank.
3. Paste the full script into the text box.
4. Use Split to divide the script by blank lines or sentence boundaries.
5. Review each phase and fix segments that are too long, too short, or incorrectly punctuated.
6. Assign a character to each phase.
7. Generate 1 to 3 phases for preview.
8. Click Generate All for batch synthesis.
9. Export or copy audio from the output directory shown in the status bar.

Split Recommendations

Problem	Recommendation
Segment is too long	Insert blank lines manually to avoid cloud context or TPM limits.
Segment is too short	Merge pure interjections or punctuation-only fragments.
Frequent character switching	Split each spoken line separately for easier voice assignment.
Narration and dialogue mixed together	Separate narration phases from dialogue phases before assigning characters.

Before Batch Generation

Check	Reason
Provider concurrency	Cloud free tiers should use low concurrency.
RPD / TPM	Avoid triggering 429 errors with long text.
Character assignment	Phases without a voice cannot be generated.
Output directory	Confirm disk space and storage path.

Character Assignment Tips

When long text contains several speakers, split it into readable short phases first, then manually assign each phase to a character from the selected voice bank.

Situation	Handling
Mostly narration	Assign all phases to the narrator first, then adjust dialogue phases.
Speaker names appear before lines	Keep names as review hints, then decide whether to remove them before generation.
One character has large emotional variation	Duplicate the character and adjust voice instruction, speed, or reference audio.
Very long batch	Generate a small set first, confirm stability, then continue.

Export Tips

• Create separate projects by chapter or scene for easier organization.
• Collect audio from the output directory after generation.
• If stable filenames matter, number phases inside the project before exporting or sorting files.