Hi all, just starting a new topic for discussion for exposing Synfig to AI Agents. I recently got a lot of exposure to AI assited coding through various tools like GitHub Copilot, Codex, Pi Coding Agent, etc. And currently I been using DeepSeek V4 Flash, and trying out different harnesses and seeing how they all work, learning from it. And I recently also developed a MCP tool that would expose clangd analysis to AI Agents, so that AI can quickly do things like Get function hierarchy, see all references, without wasting tokens on things that can be resolved statically.
Has anyone been experimenting with Synfig, and exploring using it with AI (LLMs) ? I am planning to just try out maybe create a MCP and skills to give AI capability to use Synfig, also there are open sourced alternatives like HyperFrames from HeyGen which gives ability for LLMs to generate motion graphics through web library.
Wanted to get everyones idea, do you use AI for coding, animation, etc ? What do you guys think the implication would be on animation and Synfig of AI ?
Isn’t it what any basic IDE and even Doxygen do for at least 20-30 years (without AI, Doxygen is 30yo)?
We don’t even have a decent scripting system … so AI is far
Why to even bother to add AI to a “manual tool” like Synfig while you can produce Hollywood-class 3D animation from a prompt?
The problem is not what you can do with AI to pilot Synfig (at least to generate sifz files from prompts that you can render afterwards).
The issue is about legal matters if you want to integrate generated code in the source that is copyrighted.
And also the fact that is generally poorly documented and sometimes almost impossible to modify afterwards (you can’t revert months or years of functionnalities developed as upper layers of “illegal code”), it has nothing like human planification.
I’m using exclusively Gemini because it’s free and included with Google account.
AI is practical but I could live without it.
Practical for “giving life” to ideas of funny images/memes but before (and still now) I’m using GIMP (I had plenty on Reddit before I was banned a few minutes after posting a video where I replaced a AT-AT, destroyed by Ewoks, by a Pukeko (copyright bot, account not restored) anyway Reddit is a TOXIC platform where most of users brag proudly about their mental disorders)
About coding, it permits to spare time for refactoring and producing skeletons, trivial tasks with low added value which only consume time.
But during a lot of years, I was building code generators for trivial things with Excel, like CRUD or to convert to JSON proprietary specifications of data which were on dozens of pages in PDF for to decode them from INFOLOG files …
No AI needed
Like always, a tool should stay a tool, a helper.
No to become the means of production.
Something that could help to make better and faster by freeing time and mental resources for trivial/chronophage tasks.
But not for conception/vibe coding.
More AI, less developers, hiring only the ones who master the prompt but don’t understand globally.
At the end, unsolvable problems because no one knows how it works, and global developers level will drop.
Like here in Greece, they do a lot of bragging for an “intelligent elevator” with voice recognition with AI and Lidar like if it was something new … but already existed in Japan (Mitsubishi) at least since 2011 … and no AI implied!