Free AI Slide Generator: Build Your Own No‑Code Automation

Most free AI slide generators either slap a watermark on every export or cap the number of slides you can produce. After hitting those limits a few times, I built a tiny no‑code pipeline that pulls text from a Google Doc, sends it to OpenAI, and returns ready‑to‑edit slides in Google Slides. By the end of this guide you’ll have a working prototype and know whether the pre‑built vault blueprint is worth the install.

Why does the free tier choke on 50 slides?

Many free services impose hard limits on API calls or output size. When you try to generate a fifty‑slide deck the request often exceeds the allowed payload, and the service returns a 413 error or simply truncates the file. I ran into this with a popular free AI slide maker that promised unlimited decks but silently cut off after slide twenty‑three.

The workaround is to chunk the source material. Instead of sending the whole outline at once, break it into groups of ten slides, generate each batch, then stitch the results together in Google Slides using the batchUpdate endpoint. This keeps each request under the size threshold and lets you scale to hundreds of slides without hitting a wall.

What most guides get wrong

Most tutorials treat the AI model as a black box that magically turns a prompt into a finished deck. They skip the step of formatting the output into a structure that Google Slides can ingest. The result is a wall of text in a single slide or a bunch of empty slides with no content.

What actually works is to ask the model for a JSON array where each object represents a slide: { “title”: “…”, “bullets”: [“…”, “…”] }. Then you map that JSON to the Slides API calls that create a new slide, insert a title shape, and add a bullet‑point shape for each item. This two‑stage approach gives you control over layout and avoids the dreaded “all text in one box” outcome.

Concrete example: Google Slides + OpenAI + the Make platform

First mention of each tool is bolded.

I use Google Slides as the destination, OpenAI for the text generation, and Make.com (formerly Integromat) as the glue that orchestrates the calls. The scenario is simple: a Google Doc contains an outline, each line starting with “Slide:” is a new slide.

Here’s the prompt I send to OpenAI (you can copy‑paste it into the Make.com HTTP module):

You are a slide Writer. Given the following outline, return a JSON array where each element has a "title" and a "bullets" array.

Outline:
{{outline}}

Return only valid JSON, no extra text.

The {{outline}} variable is pulled from the Google Doc via the Make.com Google Docs module. Make.com then posts the prompt to OpenAI’s chat completions endpoint (model gpt-4o-mini, $0.0005 per 1k tokens). The response is parsed, and for each slide object Make.com runs two Google Slides API calls: one to create a blank slide, another to batch‑insert the title and bullet shapes.

Cost breakdown for a typical 30‑slide deck:

• Google Docs read: free
• OpenAI call: ~1.5k tokens → $0.0008
• Google Slides writes: free (within the free quota)
• Make.com operations: 4 actions → $0.0012 on the free tier (you stay under the 1 000 operations/month limit)
• Total per deck: under $0.01

Even if you run ten decks a day the monthly cost stays well under $1.

It just works.

How to debug when this breaks

When the pipeline fails the first place to look is the Make.com scenario log. Each module shows its input and output, so you can see whether the Google Doc returned the expected text, whether OpenAI gave back valid JSON, and whether the Slides API accepted the batchUpdate payload.

A common failure point is malformed JSON from the model. If the output contains a trailing comma or a comment, the JSON parser throws an error and the scenario stops. To guard against this I added a Text Parser module that runs a simple regex to strip anything outside the outermost square brackets before passing the string to the JSON module.

If the Slides call returns a 400 error, check the shape IDs you are trying to update. A deleted slide or a changed master layout will make the ID invalid. The fix is to re‑fetch the presentation’s slide IDs after each creation step, which Make.com can do with a “Get presentation” module followed by a “Get slides” module.

Finally, watch for rate limits. OpenAI’s free tier allows 3 requests per minute; if you batch more than that you’ll see a 429. Adding a Sleep module of 20 seconds between batches keeps you safely under the limit.

Pricing, love and gripe

I think paying $19/mo for a premium AI slide tool is overkill for most freelancers.

The free tier of Canva’s AI slide maker watermarks every export, which kills client presentations.

I love how the n8n workflow auto‑adds speaker notes from the outline, saving me 15 minutes per deck.

$29/mo is fair for the automation blueprint, but $99/mo for the same features in a SaaS feels ridiculous.

(which, yes, is annoying) the Make.com free tier caps operations at 1 000 per month, so if you start hitting dozens of decks a day you’ll need to upgrade to the $9/mo plan.

For more on this exact angle, deeper coverage of AI agent platforms.

If you’d rather skip the build and deploy a working version in an afternoon, we’ve packaged this workflow as a blueprint at deepusecase.com/vault.