I track which sources AI engines cite for a living, so before writing anything I pointed the instrument at my own topic. Eight buyer questions, five engines, forty answers, and the full list of URLs those answers were built from.
Most of it went the way you would expect. Large SaaS blogs with a decade of domain authority behind them. Then something I had not planned for. All eight questions cited YouTube, 22 different videos across the set, and on the five questions that name a product directly, the same review video kept coming back.
That video has 128 views.
The part that changes what you should buy
A 128-view video sitting next to established blogs isn't a fluke. The citation research says to expect it.
- The correlation between subscriber count and AI citations is −0.03. Effectively zero. A big channel gets no advantage.
- 94% of AI citations go to long-form video. Shorts account for 5.7%.
- What does correlate is structure. Titled segments that work like headers, and transcripts an engine can pull a clean passage out of.
So the gate isn't popularity. It's format. I've written about why a 500-view review can outperform a 500,000-view video at more length, but the practical version is short. Most AI video tools optimise for Shorts, which is the 5.7% bucket. Very few produce ten-minute chapter-titled explainers, which is the 94% one.
So I made one and published it. Inside a day Google had indexed it, onto page one for adkit review at ninth in the main results, and second in the video tab.
Here's the honest reason that slot was open, and it isn't that nobody wants the traffic. You've probably thought about turning your own posts into videos and not done it. I'd been publishing reviews for four months without making a single one. Video isn't hard. It's a production. You script it, record it, edit it, render it. A post you can write on a Tuesday afternoon. A video goes on the calendar, and then it slides.
So the articles got written, the comparison pages got written, and the video never got made. There are two videos about this product. Filter almost any product category to video and it looks the same. That slot stays open until making one stops being a production, which is exactly why I went looking for a tool like this.
Two days later it was cited
I ran the same measurement again two days later, and the video came back as a source. Perplexity and Google AI Overviews both cited it, on the questions that name the product: AdKit review, AdKit review 2026, AdKit pricing. The channel has 70 subscribers, the video was 48 hours old, and the tool I'm reviewing built it out of a blog post I had already written.
Three things worth pulling out of that.
It moved fast. Eighteen hours after upload, the same forty-cell measurement found nothing. Two days later it returned six hits. That count will move around depending on the prompt and the moment, but going from invisible to cited in two days is a good sign.
It didn't replace the text. On one question the engine reached for the original blog post instead. Both are live sources now. The video added a surface rather than taking one over.
Match the video to the question. Mine reviews one product, so it landed on that product's name and skipped category questions like best Meta ads MCP. Nothing failed there. The video simply doesn't answer that question. If you want the category question, make the video that does answer it, the "so which one do I pick" comparison with your product in it.
Perplexity and Google AI Overviews are the two engines that lean on video sources, and both of them picked it up.
One thing I hadn't appreciated until I measured it: citations decay. The half-life is around 4.5 weeks. Something newer replaces you, and the slot reopens for whoever published most recently.
So the asset isn't any single video. It's the cadence, and that changes what you are shopping for. The question stops being "can this make one good video" and turns into "can this make one every few weeks, per topic, without becoming a second job." At that point the only number that matters is cost per video, and Blog2Video's is $0.60.
What it does
You give it a blog URL. It scrapes the article, splits it by subheading into scenes, writes narration, records a voiceover, pulls stock footage and renders an MP4. Every scene stays editable afterwards, including the voice track.
The voiceover is the part I keep forgetting to be impressed by, because it just happens. It runs on ElevenLabs, four voices on every plan, with premium voices and cloning if you want it to sound like you. Ten minutes of clean narration arrived without me booking anyone or opening an audio editor, and when I rewrote seven scenes the re-record took seconds. That's normally the step that kills a video project.
This is Blog2Video at blog2video.app. Check the URL, because a similarly named tool goes the other way, video to blog.
The category is crowded but mostly solving a different problem. Lumen5 and invideo compress your article into a short social cut. HeyGen and Synthesia put an AI presenter on screen. Pictory extracts key sentences instead of following your structure. None of them default to a ten-minute explainer that follows your article section by section with a title on every scene, which is the format AI search actually cites.
Time and cost, measured
| Short | Long | |
|---|---|---|
| Output | 5 scenes, 0:59 | 41 scenes, 10:14 |
| Generation | 41 sec | 4 min 34 sec |
| Render to MP4 | 3 min 22 sec | ~36 min |
Rendering runs about 3.4× the video's length. Queue it and walk away. Generation, the part where you make decisions, is the fast half.
| Plan | Monthly | Annual | Lifetime | Videos |
|---|---|---|---|---|
| Free | $0 | — | — | 1 |
| Per video | $3.99 ($2.80 at volume) | — | $299.99 / 500 | pay as you go |
| Lite | $19.99 | $15.99/mo | — | 10/mo |
| Standard | $34.99 | $27.99/mo | $999.99 once | 30/mo |
| Pro | $59.99 | $47.99/mo | $1,599.99 once | 100/mo |
Pro is where the $0.60 comes from. The 500-credit pack lands at the same number with credits that never expire, which is better if your publishing is lumpy rather than monthly.
Editing is metered separately. Regenerating a scene's voiceover costs 5 credits, anything else costs 1. I used 35 rewriting narration across seven scenes. Budget for editing, not just generation.
The free tier is one video through the full pipeline with no watermark listed, enough to judge whether the format suits your writing. Most of what's written about this tool still quotes "two free videos, $4 each." That was true once. I checked the pricing page on 2 August 2026.
Build it in the dashboard, not over the API
The least documented thing I found, and the most useful. You can drive Blog2Video from the web dashboard or over MCP from Claude, ChatGPT or n8n, and the same article produces different videos depending on which you use.
| Dashboard | MCP | |
|---|---|---|
| Stock footage | every scene | none |
| Background colour | matched to template | forced to white |
| Editing narration | re-records the voiceover | leaves the old audio in place |
Three videos I made over MCP, across two templates and two length settings, came back with no footage at all. There's no stock-footage parameter in the API call. The colour default makes it worse. MCP writes the background as #FFFFFF whatever template you asked for, and Nightfall draws titles in white because it expects a dark background. Footage would have hidden the clash. With none, the title disappears.
The third one is the quietest. Rewrite a scene's narration over the API and the text updates while the audio doesn't, so the script on record and the voice in the video end up saying different things. Neither side warns you.
Use MCP to kick off generation and check status from a chat window. Build in the dashboard.
Pick the template before you tune anything else
Fifteen built-ins, and they produce structurally different videos from the same article.
Stick Man surprised me. I expected a novelty and got a hand-drawn whiteboard explainer with my article's actual advice inside speech bubbles. "Missed one pill? Take it ASAP" on one figure, "Two or more? Use backup for 7 days" on another. That answers the real weakness in this category. Text-card templates look like slide decks because they have nothing to show but your words. Stick Man's illustration is the content.
Run your first article through two or three before settling. One generation each, under a minute apiece. The gap between the right and wrong template is wider than anything you'll fix by editing scenes later.
Custom templates get built from your website URL. It pulls your colours, fonts and tone into something reusable. Every plan includes at least one, Pro allows twenty. Above that they'll design one around your brand system, or take your articles and hand back finished videos, with pricing on request. The ladder makes sense. Start on a built-in, move to a generated custom template once you know what you want, and buy a designed one when the channel earns it.
Before you publish
Stock footage matches words literally. "The Competitor Ad Library" got footage of a man in an actual library. About a fifth of my clips landed on the wrong sense of a word. Swapping one takes seconds, but budget ten minutes for a ten-minute video.
Fix the captions. This is the step that decides whether any of it works. An AI engine can't watch your video. The caption track is the only body text it reads, which makes accuracy there the difference between being citable for a term and being invisible for it.
YouTube's auto-captions came back 93.3% word-accurate on mine, and the misses landed where they hurt. The product name I was reviewing was written wrong seventeen times. Anyone asking an engine about that product would not have found me. Twenty minutes of hand-correcting fixed it.
This is the one gap I'd most like Blog2Video to close. Every ingredient is already there. Scene narration and per-scene timings both sit in its API, and I built a working SRT out of them with a short script. An export button would remove the step, and given how much of this product's value rides on getting found, it's the highest-leverage thing they could ship.
Who it's for
Worth it if you publish written content regularly and want a video surface without hiring anyone. Software reviewers, SaaS blogs, newsletter writers, documentation teams.
Wait if your video has to show your product moving. There's no screen recording and no footage upload, so anything you want on screen has to be an image in the article first.
Verdict
4 / 5. It turns writing you've already done into structured long-form video, faster and cheaper than anything comparable. Build in the dashboard, test two templates, skim the preview for invented data, fix the captions.
I didn't set out to prove it would work, only to find out whether it would. The answer is a qualified yes. A video this tool made from an article I had already published became a cited source in two AI engines within two days, on a 70-subscriber channel. How long that holds is still open, and at $0.60 a video you can find out the only way that works, by publishing the next one.
Based on three videos generated and rendered on 1–2 August 2026, one of them the full 41-scene version. Pricing checked 2 August 2026. Citations measured across eight buyer questions on five AI engines, before publication and again two days after.