01
What are these policies aimed at?
Synthetic depiction. A generated face, a voice that was never spoken, a scene that never occurred. Labelling obligations attach to material that could mislead a viewer about whether something real happened. Editing genuine footage, including choosing moments and adding words that were actually said, sits outside that.
02
Generated against edited
These policies target synthetic depiction, meaning material that could mislead a viewer about whether something real happened.
| What was done | Generated? | Labelling generally required? |
|---|---|---|
| Cutting a real recording | No | No |
| Cropping and adding words that were spoken | No | No |
| A synthetic voice reading real text | Yes | Yes |
| A generated face or person | Yes | Yes |
| A scene that never happened | Yes | Yes |
| Colour and sound adjustment | No | No |
03
Cases that sit in the middle
Some uses fall between editing and generation. Replacing a background, cleaning up a voice, or translating speech with a synthetic voice may or may not trigger labelling rules, depending on the platform and how realistic the result is. When a change could make a viewer believe something happened that did not, lean toward labelling.
Read each platform's current guidance, because examples and thresholds are updated often.
04
Using labels without losing trust
A clear label, applied honestly, rarely hurts trust with viewers. What hurts is discovering later that something was generated without being told. If you use generated elements, a short note explaining what was created and why can turn a potential problem into transparency.
Platform labelling tools are usually simple toggles at upload time; use them when the rules call for it.
05
Keeping records of what you made
Keep notes on which parts of a clip were generated or altered and with what tool. If a platform or viewer questions a clip, you can explain quickly and accurately.
This is especially useful for sponsored content, where brands may also need to know.
06
Editing real footage
Cutting, reframing, adding subtitles and adjusting colour on real recordings is ordinary editing, the same as any video editor has always done. The footage still shows what happened. That kind of work is not what synthetic-media rules are aimed at.
Keep your original recordings, which prove the footage is real if ever needed.
07
Explaining your process to viewers
Some viewers are curious or suspicious about how clips are made. A short, plain explanation in your profile or a pinned post, such as clips cut from our full conversations, builds trust and answers questions before they are asked.
Keep the explanation about what viewers get, real conversations turned into short clips, rather than technical detail. People want reassurance that what they see is genuine, not a tour of the workflow.
08
Label what was fabricated
- Ask whether anything depicted is fabricated. That is the question the rules turn on.
- Label synthetic voice or likeness. Both are explicitly covered nearly everywhere.
- Do not over label real footage. Marking genuine material as synthetic misleads in the other direction.
Clip real footage and post it plainly.
Questions people ask
- Does editing real footage need a label?
- Generally no. Cutting, cropping and adding words that were spoken is editing rather than generation.
- What about a synthetic voice reading real text?
- That is a generated element and is covered by labelling rules in most current policies.
- Are these rules settled?
- No. This is the fastest moving area of platform policy, and anything specific is worth rechecking.
- Does the label reduce reach?
- There is no reliable evidence either way. Failing to label where it was required has clearer consequences.
- Does jurisdiction come into it?
- Increasingly, as several jurisdictions legislate on synthetic media independently of platform policy.