Audio-to-Audio & Strength Tuning

When using audio-to-audio workflows, the following are recommendations for settings:

Longer audio input tends to improve the output based on length (for instance, a minimum of 6 seconds audio input for a 180 second output will produce less consistent results than a longer input, say 190 seconds).

Adjusting the "strength" parameter will vary the degree to which an input prompt changes the input audio. A recommended value of 0.8 is used for starting, and can be increased for greater variation, or decreased for increased consistency with the input audio. Gradual adjustment is encouraged here.

Prompts are encouraged to be descriptive, rather than instructive. For example:- Use descriptive prompts
- Avoid command-style phrasing (eg "Make this a 70's style rock track" in place of "vinyl recorded acoustic guitar with soft drums, 4/4 time signature").
- Avoid negative exclusions (eg "no reverb" may still include reverb).
- Expect experimentation.

Strength Guidelines

| Goal | Strength |
|------|----------|
| More variation | 0.85–0.95 |
| More adherence | 0.6–0.75 |