Speech, music and transcription
Keep narration, voice cloning, sound design and captions as distinct operations.
Voiceover and voice changes
Use /rikoko /speech for standalone narration or later voice replacement. The current command default is Eleven v3 when its profile is available. Speech uses the tts operation with explicit model, text, language and voice. Keep these settings out of video-generation parameters. Preserve native video audio unless the requested edit calls for a replacement.
Clone and reuse a voice
The minimax-voice-clone example requires an authorized voice recording of 10–60 seconds, up to 20 MiB, bound with the voice role. Save the resulting voice ID and clone asset. minimax-cloned-speech requires that voice profile for reuse; it is not a text-only default narrator. A short, clean recording without overlapping speakers helps isolate the intended voice.
Music and sound effects
Use /rikoko /audio and select the correct category. The example elevenlabs-music-30s profile produces a 30-second instrumental bed; elevenlabs-sound-5s produces a five-second effect. Other durations require supported overrides or separately configured profiles and pricing. Specify the mood, instruments or sound source, then position the result on the edit timeline and balance it against speech.
Transcription and caption timing
Scribe v2 is a transcription profile, not a speech generator. Retrieve the completed transcript through transcript_get and select actual words or source intervals with transcript_select. Build captions from returned word timings rather than estimating them from text length. Check names, numbers, speaker changes and pauses against the audio before final export.