Synthetic media, an umbrella with soft edges
Synthetic media is the umbrella term for material a model produced or changed. It is useful for naming a subject and useless for deciding whether a particular deliverable is inside it. As of 2026-09-12.
| Case | Clearly inside the term |
|---|---|
| A clip generated from a prompt | Yes |
| A synthetic voice reading a script | Yes |
| Shot footage with a generated background | Partly, and nobody says which part |
| Shot footage graded with a model | Unclear, and this is the common case |
Inclusion rule. Production cases against the umbrella term as instruments and commentary use it. Order. From the clearest case to the least clear.
1The edge is where all the real work happens
Fully generated material is easy to classify and rare in professional delivery. Almost everything shipped is a mixture, and the interesting question is how much model involvement makes a deliverable synthetic.
Instruments in this area tend to gesture at the problem with a phrase about substantial alteration and leave the line undrawn. That is honest and it is not guidance.
2Assistive editing is the carve-out that everybody reaches for
Transparency instruments frequently exempt standard editing that does not substantially change the input. Every modern editor now contains model-driven tools, which makes the exemption both essential and vague.
A production cannot resolve that from documentation. What it can do is record what was used on what, so that whoever eventually has to decide is deciding from facts rather than from recollection.
3The term is a subject, not a test
Used to name a field it is fine. Used as a classification for a specific file it produces arguments, because the two participants are applying different implicit thresholds.
Which is why registers organised around this subject work better when they record mechanisms and statements rather than classifying material. Classification is the part that needs a decision-maker.
4Audio and text sit inside it and get discussed least
Generated speech and generated text are named in the same instruments as video and attract a fraction of the attention. Neither has a visible layer, so both depend entirely on records and disclosures.
A production delivering a narrated piece is often further inside this umbrella than it realises, and with fewer tool-side options than a picture would have.
5Published material behind this entry
A term in this record: what it means, and what the published material says about it, each figure read on 2026-09-12. Related: Open weights, Test mode.
6A provider's own account
Against the duty in “Synthetic media”, SceneMixer's published pages describe what its pipeline produces:
SceneMixer writes dialogue into the script analysis, assigns each line to a shot with a speaker and a short delivery note, and the video model performs it — picture and audio generated together, so the mouth forms that sentence in the language you chose.
SceneMixer, AI Dialogue Generator (read 2026-10-13)
Dubbing replaces the audio of a finished clip; the picture was generated for a different line, so mouth shapes, pauses and shot length all belong to the original language. Native-language generation gives the video model the line in the target language before it renders the shot, so the performance, the pacing and the cut points are built for that language.
SceneMixer, Native-Language Dialogue vs. Dubbing in AI Video (2026): Why the Cast Should Speak the Line (read 2026-10-13)