Talk:Synthetic media: Difference between revisions

Wiki Education Foundation-supported course assignment

This article was the subject of a Wiki Education Foundation-supported course assignment, between 24 August 2020 and 16 October 2020. Further details are available on the course page. Student editor(s): Willland, Awchesley.

Above undated message substituted from Template:Dashboard.wikiedu.org assignment by PrimeBOT (talk) 04:00, 18 January 2022 (UTC)

Possible source

Not sure if you have found this source, but you may find it useful in your drafting.

Also, while I am here, I don't think you need double or triple citations. When you have a claim, one good citation provides verifiability to the claim. If you have lots of sources you wish to refer to, consider adding them instead as links to articles for further reading. -- Sirfurboy (talk) 08:26, 4 February 2020 (UTC)

My problem with this is that some extraordinary claims are only supported by a single citation. This article is establishing a whole field of media and research. You would expect hundreds of available citations to support core facts, like the term use or what it means. Instead, what is cited is all that is available and the citations don't even agree with each other or the name. And once you go past the lead, the citations don't even support a connection to "synthetic media". I really don't have the time to go sentence by sentence to verify that cited content is indeed exactly what the sources say. —  HELLKNOWZ   ▎TALK 10:46, 4 February 2020 (UTC)

The fundamental issue I (and a couple others) have recognized is that the use of "synthetic media" specifically as a description of this field is quite literally no more than two or so years old in mainstream parlance and was enabled by deepfakes. Before roughly 2018, discussion of synthetic media was rare, and there was no one, hard-established term for it, except arguably "AI-generated media" (which is still used interchangeably with "synthetic media"). Starting around 2018, more media pieces and research groups began honing in on "synthetic media" as the catch-all term, and this is why you can find constant references to synthetic media = AI-generated media, deepfakes, image synthesis, music synthesis, and so on but it becomes much more difficult to actually find papers and articles that actually go in depth with this. It's not quite as scarce as it was even a few months ago since "synthetic media" is now rapidly becoming a much more widely discussed field, but as HELLKNOWZ has mentioned, you have all these references and leads that outright say that "image synthesis is a form of synthetic media" but then barely use the term going forward in the articles or papers themselves, often because, again, it's still used interchangeably with "AI-generated" media or reduced to "image synthesis". The issue with this article is that it's being drafted two years too soon, before more numerous in-depth explanations of synthetic media could be made. And yet even then, as Sirfurboy has mentioned, there are more sources being written to support/reinforce the article, often even going in-depth. In fact, just today I discovered even more sources that define synthetic media which I planned to add to the main article:

• https://www.youtube.com/watch?v=1UAVjBIs9gs
• Twitter's ban/regulation of synthetic media
• Another mention by Commonsense.org that deepfakes are a type of synthetic media
• Witness.org doing a bit more in-depth run-down of synthetic media, but again referencing the sub-branches
• Towards Data Science's "Fake New World" piece

It defines synthetic media yet again: "The idea behind SM is that an AI can be used to generate images, text, voice, videos and practically any other media."

It's frustrating because these cited articles do support these claims, but not in high-enough volume to do so comfortably, and because "synthetic media" is indeed always used as a starter with sub-branches then used going forward. — Yuli Ban (talk) 04:57, 7 February 2020 (UTC)

Maybe take a look at WP:THREE and see what the best three sources you can find are. It is not the volume of sources you need - just three good ones. However, if there are not even three good sources then maybe you are right about this being written too soon. Wikipedia is a tertiary source, so we have to wait for a subject to be discussed in some secondary sources before we can say it is notable. If I see anything else I will definitely let you know. -- Sirfurboy (talk) 07:26, 7 February 2020 (UTC)

The three articles I've found that best exemplify synthetic media are:

• https://www.synthesia.io/blog-post/the-future-of-synthetic-media
• https://securingdemocracy.gmfus.org/deepfake-myths-common-misconceptions-about-synthetic-media/
• https://blog.paperspace.com/2020-guide-to-synthetic-media/

I've been reading the news lately, and it's become apparent that this subject has become urgent. I'll relent completely and delete whatever parts of the article are causing hiccups, just so it can go live as soon as possible and more people can become aware of the term. — Yuli Ban (talk) 18:22, 12 February 2020 (UTC)

Synthesia article is not independent, the company in question is directly engaged in the field. —  HELLKNOWZ   ▎TALK 18:29, 12 February 2020 (UTC)

I see. There are two other good sources that might be able to replace it, the first being an academic paper.

• https://arxiv.org/abs/1907.11274
• https://www.newyorker.com/magazine/2018/11/12/in-the-age-of-ai-is-seeing-still-believing — Yuli Ban (talk) 20:20, 12 February 2020 (UTC)

adding a music example

Should I attach an audio file sample (MP3) from Riffusion? It generates music from text using similar technology to Stable Diffusion. The music sample would be seven seconds long. Camdoodlebop (talk) 00:40, 18 December 2022 (UTC)