Synthetic Narration
Choose Quick TTS, HKGAI, or Qwen according to the voice you actually need.
Write for the ear
A readable paragraph is not automatically speakable. Short sentences, clear punctuation, and a pronunciation check make one synthetic take easier to edit into a shot.
- Draft a single approved line, for example “The lights come on. The studio is ready.”; read it aloud and mark a natural pause.
- Place Audio → Text-to-Speech on the canvas, or connect a Text node when you want to maintain an editable script upstream.
Decide whether Quick TTS is enough
The menu’s Text-to-Speech tool starts as a Quick TTS Audio node. The repository also provides a 302AI TTS route; a built-in quick voice is not the same thing as an enrolled personal voice.
- Select the Audio node, enter the line, choose an available voice or speech settings, and run a test.
- Preview intelligibility and accent before connecting the Audio result to a speaking portrait or an edit.
Use HKGAI advanced speech when needed
HKGAI TTS has Text, Built-in voice, Voice ID, Instructions, and Reference language controls. The editor can use connected upstream text when its own text field is empty.
- Switch the TTS type to Advanced TTS (HKGAI) if the advanced panel is available; choose a built-in voice such as a Mandarin or English option.
- Enter “warm, natural, conversational” in Instructions, synthesize, then replay names and numbers carefully.
Use Qwen for an enrolled voice
Qwen Voice TTS offers a cloned-voice selector, Target model, and Language type. Its Voice input carries an enrolled Qwen voice ID, whereas HKGAI TTS cannot consume that ID.
- Connect an authorized Voice Cloning result to Qwen Voice TTS’s Voice port, then enter or connect the script to Text.
- Select the enrolled voice, review Language type and the text-length indicator, and synthesize a short test before a whole narration.
Keep the audio usable downstream
A completed audio node exposes an audio player; some provider result links are temporary, so playback now does not guarantee the same remote link stays valid indefinitely.
- Download or archive the approved take as permitted, then connect its Audio output to Video Edit or a compatible digital-human Video input.
- If the next node cannot find sound, inspect the audio result and connection type before regenerating text or changing models.