Specific topics
GenAI - best practice to minimize risks
1 |
2 |
3 |
4 |
5 |
6 |
Guidelines of the respective Sonova brand must always be taken into account |
Ensure the AI-generated sound doesn’t mimic or derive from copyrighted/trademarked sounds.
|
Use AI sound tools that provide clear, commercial-use licenses. |
Keep records of how the AI-generated sound was created in case of a dispute. |
Do not use free AI models with unclear data sources |
For YouTube, test sounds before publishing to avoid Content ID flags |
|
Styles, sounds and voices are often already regulated in the corresponding brand guidelines. Compliance with these always has top priority. |
|
Do not use comparative prompts (e.g. “Sounds like “Lady Ga Ga”) , “similar to Beyoncé’s …” “could be sung by Bruno Mars”, ‘like the sound of “The Muse”, etc.). Listen to the sound and assess short musical phrasing of a couple of seconds can be protected. If the music sounds familiar, do not use. |
|
See the list provided for potential sources. |
|
Archive all relevant information, incl. prompt to generate the relevant sound file. |
|
Risk of unlicensed training. |
|
YouTube has a strict copyright enforcement system, primarily using Content ID, an automated tool that scans videos for copyrighted audio. If a sound does not pass Content ID, it’s an indication it’s likely copyrighted : do not use that sound anywhere. |