How to Ensure Template Formatting and Design Consistency When Generating with Gamma API | Gamma Community

Gamma Community Icon

How to Ensure Template Formatting and Design Consistency When Generating with Gamma API | Gamma Community

Max J.
·
·
Hey William 👋 Thanks for the detailed writeup and the screenshots, those are exactly the kind of before/after comparisons that make it way easier for us to pinpoint what's going wrong. Really appreciate you taking the time to document it all. On the overlay and effects issue (superposition converting to blur) and the header formatting loss where spans are getting stripped out - these are known issues we've been actively working on. Our team has fixes in progress, so if you try again tomorrow you should see improvements on both of those. I know it's frustrating when things shift between what you designed and what comes out the other side, so thanks for bearing with us while we get these dialed in. For the card-level settings like fullscreen mode and width, and component-level params like logo grid spacing - those aren't currently controllable at the prompt level, so there's no workaround I can offer on that front right now. We are tracking it though, and your report adds to the case for prioritizing it. One thing worth flagging: if you have flexibility on which endpoint you're using, cardOptions is available on the regular /generations endpoint. That might give you more control over some of the card-level behavior depending on your use case. If you're still seeing issues after tomorrow's fixes go out, drop a before/after comparison here and I'll dig into the specifics with you. The more examples we have the faster we can nail down edge cases. Thanks again for helping us make this better!
Nikolas P.
·
·
Thanks William. Super helpful 🙏 Max J. did you cut a tix for this?
William
·
·
Thanks a lot for your answer Max and Nik, greatly appreciating you taking the time and being so reactive ! To give you a bit more context on the use case, it's really a ingest and summarize a lot of prospect context via LLM -> swap in some context only on some slides only (like 6 cards over 25), so the generations route doesn't really fit the need. I feel like for this kind of use cases, the from-template route might benefit from duplicating the template and working from there, editing instead of regenerating from scratch, which directly leads to the formatting issue mentioned earlier or the /from-template might benefit from getting the html version of the gamma, which would include span and custom formatting etc, allowing the agent to rewrite from scratch while keeping formatting intact just my 2 cents here 🙂 I think what you introduced with remix and are consolidating with from-template is extremely powerful, especially if you find a way to ensure a pixel perfect design consistency, so using gamma doesn't become a design break hunt, which sadly makes it impossible to deploy programmatically at scale if we need to keep a human in the loop to review and fix manually each card's component settings.
Nikolas P.
·
·
Jens W.
Jens W.
·
·
Thanks for sharing this feedback! Would be great if you can send the doc that you've generated (or share the ID) so I can investigate what happened here
Nikolas P.
·
·
William a fix for this should be going out on Monday
William
·
·
Hey Max J. Nikolas P. Jens W. ! Quick message to let you know this has been fixed ! the from-template doesn't break fine formatting like span, and now respects each component settings from the template (e.g a logo grid keeping its template's spacing and logo values) a life changer, no more manual review of generated gammas. Thanks a lot for your reactivity, it is a pleasure to build with gamma, and looking forward to seeing what you guys are cooking :)
💙1