Gamma Studio: Create AI Visual Stories in Minutes (Hands-On Demo)

Full video transcript

How’s it going everybody? Darby here with GenAI University, tuning in the day after the Scale With AI virtual summit, joined today by members of the paid community for a live interactive workshop specifically going into Gamma and implementing some of what was learned yesterday. The purpose of this workshop is implementation of what was learned yesterday and how to practically use Gamma inside of visual content creation and marketing. This session covers some of the latest Gamma features and functions that have been rolling out — some of which may not be available in every account yet, since this is a brand-new feature called Studio in Gamma. It came out less than 24 hours before the summit went live, and the Gamma team hadn’t even announced it yet; it was spotted via a post inside Gamma’s Game Ambassador community.

Inside Gamma, the presenter walked through an existing project — 60 slides generated in one pass, following a script and format provided as an upload. The color scheme, words, and images were all AI-generated, not Gamma’s traditional editable card format. Going into a project’s settings shows the source prompt that created the presentation. The example was described as “making a commercial that feels like a trailer/luxury billboard campaign,” using a prompt drafted with GPT-5’s help, broken into individual slides with instructions for each. Settings include minimal-text-with-outline, a visual theme (including custom brand themes with palette, colors, and typography), and an image art style — photorealistic, a Gamma-branded 3D look, or a fully custom style built from an uploaded brand style guide.

Studio can generate not just presentations but social media content too. The presenter has a few recurring content needs — GenAI University, client work using the “market sauce” method, and the Side Hustle party game — and demoed Side Hustle next to show a different use case: an onboarding deck for a birthday-edition version of the game, meant to teach a facilitator, teacher, or parent how to run it (game objective, how to win, team setup, and so on).

Editing inside Studio doesn’t work like traditional Gamma text editing, since everything is image-based. When a word (the crown reading “investor”) was misspelled, the presenter typed a plain-language instruction to fix the spelling and clean up the letter shape. The first attempt didn’t apply; a repeated instruction fixed the spelling correctly, though the color fill wasn’t fully corrected either. The presenter compared this interaction style to prompting vibe-coding tools like Bolt or Lovable, and to Descript’s AI “Underlord” chat-based editing.

To build the 60-slide version, the presenter had set up a dedicated ChatGPT project already loaded with the brand style guide and full game instructions, then generated the storyboard narrative prompt from there. For a second pass, they used voice dictation (Whisper Flow) to ask for a more condensed 15-slide version aimed at facilitators, with more text per slide and clearer creative direction, plus a shorter “prompt style guide” to paste into Gamma. That output correctly used the brand’s custom theme colors and largely nailed the intended game components, cards, and steps, with only minor issues like an out-of-order step.

The presenter then tried the same brand content with a photorealistic art style instead of the custom illustrated one. Photorealistic human characters were consistently flagged in the workshop as unsettling (“creepy,” “horror game,” “giving kids nightmares”), and the group repeatedly steered back toward illustrated, whiteboard-doodle-style output instead. When the mascot (a unicorn wearing sunglasses, tied to the game’s Kickstarter-era branding) didn’t reliably appear correctly across styles, the presenter refined the custom style prompt to explicitly describe the mascot’s sunglasses and color scheme, which fixed the inconsistency.

Beyond the slide deck, the presenter demoed a companion AI chat tool (Pickaxe) built to help a facilitator run the game live — asking about player count, time available, and age group, then generating a tailored round structure, team sizes, and random business-idea prompts pulled from a larger list of names and industries. The chat tool supported voice input and was configured with instructions, tone, word biases, and a creativity setting, and had been tested against both GPT-5 and Claude 4 to compare response behavior.

Later in the session, Studio was tested for two more use cases: generating three YouTube thumbnail variations per topic from a full set of summit session transcripts (with a workshop discussion on why outcome-driven, non-clickbait titles tend to perform better long-term for an educational channel), and generating square social-media quote graphics — short, on-brand quotes paired with illustrated visuals, refined slide-by-slide with plain-language chat edits like removing extra text under a headline. A guest on the call, John, added context from a separate synthetic-video/ad project: static image ads were reportedly outperforming AI video ads for some of the accounts he was testing, reinforcing the workshop’s interest in fast-iterating image-based content.

The presenter closed by noting Gamma Studio was, as of the recording, available only to members of Gamma’s Game Ambassador community, that Gamma had not confirmed whether or when it would roll out more broadly, and encouraged viewers interested in Gamma to look into joining that community. The session ended with a reminder to check out Gamma for presentations, websites, social media, and documents, and to subscribe to the channel for future workshops.