How to Create an Educational Video with AI: Synthesia and ChatGPT, Step by Step

By the end of this tutorial, you’ll have a two-to-three-minute educational video presented by an AI avatar, with no camera or filming: ChatGPT writes a script split into scenes, and Synthesia has an avatar deliver it in the language you choose. It takes six steps. You need a ChatGPT account (the free plan is enough) and a Synthesia account. Synthesia’s free plan includes 10 minutes of video a month but no MP4 download; downloads start on Starter at $29/month, or $18/month billed yearly (Synthesia pricing, checked Sept. 25, 2026).

We rewrote this tutorial on September 25, 2026, based on the Synthesia Help Center and the official pricing pages. Since the 2025 version, Synthesia has added a built-in Assistant that drafts and lays out a video from a prompt or a document, the free plan doesn’t include downloads, manual gestures have given way to avatars whose expressions follow the tone of the script, and ChatGPT’s models have changed. Button names below match the Help Center.

What You Need and What It Costs

• A desktop or laptop computer: the Synthesia editor isn’t available on mobile.
• A topic and a specific learning objective: “viewers can calculate a percentage” beats “percentages.”
• A ChatGPT account: the free plan is enough to write and review a script. Paid plans (Go at $8/month, Plus at $20/month) mostly raise usage limits (ChatGPT Go and ChatGPT Plus help articles, US launch pricing, checked Sept. 25, 2026).
• A Synthesia account: which plan you need depends on how you’ll deliver the video (see the table below).

Which Synthesia Plan Do You Need for Educational Videos?

It all comes down to delivery: a share link, an MP4 file, an interactive video or a tracked module in a learning management system (LMS). US prices from the Synthesia pricing page; features checked in the Help Center (downloads, interactivity, quizzes), all on Sept. 25, 2026.

A video with buttons or branching can’t be downloaded as an MP4: it only works in Synthesia’s player (link, embed or SCORM). As for quotas, each generated second costs 2 credits, or 120 credits a minute, and the free plan includes 1,200 credits a month (Synthesia credits, checked Sept. 25, 2026). Previews are free. Budget example: a 3-minute video uses 360 credits. That leaves 840 of monthly Starter’s 1,200 credits: 7 more minutes of video or 33.6 minutes of dubbing without lip sync. Both draw on one balance; you do not get 10 minutes of video plus 48 minutes of dubbing. Calculated from the official credit costs, checked Sept. 28, 2026. For plans, avatars and languages in detail, see our Synthesia guide.

Step 1: Write the Script with ChatGPT

Open ChatGPT and paste this prompt, filling in the brackets. It asks for what an avatar needs: text written to be spoken, already split into scenes, with what should appear on screen.

You are an instructional designer. Write the script for a 2-minute educational video (about 280 words) on [TOPIC] for [AUDIENCE: 8th graders, new hires, complete beginners...].
Learning objective: by the end, viewers can [SPECIFIC, CHECKABLE OBJECTIVE].

Constraints:
- An AI avatar will read this aloud: sentences of 8 to 16 words, no bullet points in the spoken text, no abbreviations, numbers written the way you'd say them.
- Split it into 6 numbered scenes: scene 1 = a hook phrased as a question; scenes 2 to 5 = one idea each, with a concrete example; scene 6 = a three-point recap and one action to take right away.
- For each scene, give the spoken text first, then an "On screen:" line with a title of 6 words max and a visual idea.
- Tone: [clear and friendly / professional].

Why 280 words for two minutes: Synthesia recommends pacing at 130 to 150 words per minute and keeping sentences between 8 and 16 words for talking-head videos (Synthesia’s script-writing guide, checked Sept. 25, 2026). For three minutes, aim for 400 to 450 words. For more on writing with ChatGPT, see our tutorial on writing high-quality content with ChatGPT.

Step 2: Have the Script Reviewed for Narration

ChatGPT can get facts wrong. In the same chat, ask for a critical review, then check every flagged point yourself.

Review this script like a demanding instructor:
1. Flag any claim that is inaccurate, questionable or too vague, and say how to verify it.
2. List acronyms, proper names and foreign words a synthetic voice is likely to mispronounce, and suggest a phonetic spelling.
3. Split any sentence longer than 16 words.
4. Estimate each scene's length at 140 words per minute.
Then return the corrected script in the same format.

Keep the list of tricky words: you’ll fix their pronunciation in Synthesia in Step 4.

Step 3: Create the Video in Synthesia

Sign up for Synthesia on a computer (affiliate link: we earn a commission if you subscribe, at no extra cost to you). There are three ways to start from your script, depending on your plan.

Any Plan: Start from a Template, One Scene per Section

1. On the Home screen, open More, then Explore templates, and pick a template (60+ on every plan).
2. In each scene, click into the script box and paste that scene’s text. In Synthesia, each scene has its own script, and scene breaks act as natural pauses (script help).
3. Add any missing scenes with Add scene in the canvas toolbar, then replace the on-screen titles with the “On screen” lines from your script.

Any Plan: Import a PowerPoint Deck

If you already have slides, this is the fastest route: with Import PowerPoint on the Home screen, speaker notes become each scene’s script, and Synthesia writes one from the slide text when the notes are empty. It takes .pptx files up to 150 slides or 1 GB; animations aren’t imported (PowerPoint import help, checked Sept. 25, 2026). To prepare the slides, ask ChatGPT:

Turn this script into a slide outline. For each scene: a title of 6 words max, at most 2 short bullet points to show on screen, and the spoken text to paste as-is into the speaker notes.

Starter and Up: Synthesia’s Assistant

The built-in Assistant (Starter, Creator and Enterprise) doesn’t use credits (Assistant help, checked Sept. 25, 2026):

• To keep your text word for word: on the Home screen, select More, then Import script, and paste your script. The Assistant uses it without rewriting it.
• To start from a document: select Create video, describe the topic, audience and objective, attach your script or source material (PDF, PowerPoint, Word or a URL) with the + button, and choose a length (Short, Medium or Long).

Either way, pick a delivery style (Presentation or Dynamic) and a template (Auto lets the Assistant choose). If the Assistant proposes an outline, review it, then choose Continue to editor rather than Generate video so you can check everything before generating.

Step 4: Choose the Avatar, Voice and Pronunciation

1. Avatar: the Avatar tab at the top of the editor. To swap an avatar everywhere, pick a new one and confirm with Change All (avatar help). The free plan includes 9 avatars; Starter includes 125+.
2. Language and voice: Synthesia detects the script’s language. To change the voice, click the speaker pill in the script box, then Change Voice, and filter by language or accent. Available voices depend on the avatar, so choose the avatar first (language help).
3. Pace: hover over the speaker pill and set the speed between 0.8x and 1.2x; slow down the harder parts (voice help).
4. Pronunciation: highlight a word, select Pronunciation, type a phonetic spelling or record yourself saying it, then select Use (pronunciation help). This is where the word list from Step 2 pays off.
5. Pauses: put your cursor in the script and add a pause with the pause button on the timeline toolbar (1 second by default), for example after asking viewers a question.

Avatars now run on the Express-3 model, whose expressions follow the tone of the script (Express-3 announcement). On an expressive avatar, gestures are generated from the text and can’t be turned off; punctuation and questions shape the delivery (expressive avatar help).

Step 5: Visuals, Captions and Preview

• One visual per idea: add images, video clips and shapes from the media library (more limited on the free plan). Synthesia suggests changing visuals at least every 30 to 45 seconds (script-writing guide); its French-language article on training videos goes further and aims for a new scene every 10 to 20 seconds.
• Captions: closed captions that viewers can toggle are included by default and match the script. For styled on-screen captions, open Captions in the editor’s top menu (captions help).
• Preview: the play button in the script box plays one scene; the one next to Generate plays the whole video. Previews don’t use credits, but the avatar doesn’t move in them. That’s expected: animation only appears in the generated video (preview help).

Step 6: Generate, Share or Download

1. Select Generate at the top right, add a title and, if you want captions burned into the picture, turn on Burned-in captions. The video goes through moderation, then rendering (generation help). This is when credits are deducted.
2. Share (any plan): select Share, then Create link. Copy the link or an embed code; an embedded video updates automatically when you edit it in Synthesia (sharing help).
3. Download (Starter and up): open the video’s ⋯ menu, then Download: MP4 in 1080p, audio only, or captions as .srt and .vtt files.

Add a Quiz to Your Video

ChatGPT can draft the questions in seconds:

Based on this script, write 4 multiple-choice questions to check understanding. One correct answer per question, 3 plausible wrong answers, and a one-sentence explanation of the correct answer. Only ask questions whose answers are in the script.

Where they go depends on your plan. Quizzes built into the video (single- or multiple-answer questions, scoring, a pass mark) are Enterprise-only, like SCORM export. On Creator, a button can send viewers to another scene or a link. On other plans, put the questions in your LMS quiz or in a form right after the video.

Common Mistakes and Fixes

• A script written to be read, not heard: long sentences, lists and parentheses sound off when an avatar says them. Rerun Step 2.
• Mispronounced acronyms: acronyms may be read as a word instead of letter by letter, or the other way around. Fix them with Pronunciation; to have letters read one by one, type them in lowercase with spaces.
• No download button: three documented causes. You’re on the free plan; the video contains interactive elements; or the last scene isn’t set to end the video. In that last case, fix the final scene’s transition and regenerate (download help).
• Out of credits: once your credits are used up, usage is capped until renewal unless you upgrade or buy extra credits, and unused minutes don’t roll over. Preview as much as you need before generating.
• A scene that runs too long: a single scene’s script can’t exceed 5 minutes. For an educational video, one idea per scene is enough.
• Factual errors: the avatar will faithfully repeat a ChatGPT mistake. Have someone who knows the subject review the script before you generate.

Alternatives to Synthesia for Educational Videos

• Colossyan: consider it for interactive training. Pricing lists five monthly streaming SCORM exports on Professional, but the SCORM guide still restricts Video exports to Enterprise (checked Sept. 30, 2026). Confirm the subscription and validate a sample package in your LMS; the Synthesia vs. Colossyan comparison provides a checklist.
• HeyGen: geared more toward marketing and social media. The free plan is limited to 3 one-minute videos a month with a watermark; Creator costs $29/month or $24/month billed yearly, with videos up to 30 minutes (HeyGen pricing, checked Sept. 25, 2026).

For a complete course rather than a single video, see the Coursebox profile, which explains the conditions for dynamic and self-contained SCORM exports.

If you’d rather illustrate a lesson with generated footage than a presenter, see our guide to the top AI video tools in 2026.

FAQ

Can I make an educational video for free with Synthesia and ChatGPT?

Yes, with free ChatGPT and Synthesia’s Basic plan: 10 minutes of video a month and 9 avatars. The video keeps the Synthesia logo and can only be shared by link or embed; to download the MP4, you need at least Starter (pricing, checked Sept. 25, 2026).

Do I need ChatGPT Plus to write the script?

No. A script of a few hundred words and its review fit easily within free ChatGPT. Paid plans mostly raise limits on usage, file uploads and image generation.

How many words is a one-minute video?

About 130 to 150 words, according to Synthesia’s script-writing guide. A two-minute video is roughly 260 to 300 words. You can then adjust the voice speed between 0.8x and 1.2x.

How long should an educational video be?

Short. Synthesia recommends no more than 6 to 8 minutes, noting that engagement tends to drop for longer videos in workplace research (training video guide); its French-language article, citing research on online educational videos, puts peak engagement at six minutes or less. Several short videos, one per objective, work better than one long module.

Can Synthesia add quizzes or export SCORM?

Yes, but only on Enterprise. Creator adds buttons and branching. On other plans, put your questions in your LMS or a form after the video.

Can the avatar speak languages other than English?

Yes. Synthesia offers 160+ languages and voices on every plan and detects the script’s language to suggest matching voices.