InVideo vs Revid vs ViewMade — the six measurements that decide it

We are one of the three products in this title, which is a reason to distrust the article, so let us set the terms before anything else.

We have not run InVideo and Revid through our own measurement pipeline. Publishing scored output for competitors we did not measure would be exactly the kind of comparison this category is already full of. What follows is therefore two things: the structural differences that are verifiable without a subscription, and the six measurements that let you settle it yourself on your own subject in about twenty minutes.

If you want a ranked list with a winner, this is not it, and any article offering one has either been paid or has tested one product.

The structural difference

The three tools answer different starting questions, and this decides more than any feature.

InVideo starts from a prompt or a script and assembles from a stock library plus generation. It is the broadest of the three: many formats, many aspect ratios, general-purpose.

Revid starts from a prompt and leans on generated visuals with stock alongside. It is built around speed and volume for short-form social output.

ViewMade starts from a topic and sources real archive footage, then ships a media credits file naming every clip used. It is the narrowest of the three by design — vertical documentary Shorts — and the narrowness is the trade.

	InVideo	Revid	ViewMade
Starts from	Prompt or script	Prompt	Topic
Primary visual source	Stock library + generation	Generation + stock	Real archive footage
Names its sources	No	No	Yes, per render
Output breadth	Many formats	Short-form social	Vertical documentary Shorts
Uploads for you	Scheduling available	Scheduling available	No — read-only channel access, you publish

Two of those rows deserve more than a checkmark.

Source naming matters when a video makes a claim about a real event. "Which recording is this" either has an answer or it does not, and a pipeline that cannot produce one has told you something about how it assembled the video.

Upload access is the row people skim and should not. Scheduling requires upload permission on your channel, which is a much larger grant than reading its statistics. That may be a fair trade for the convenience; it should be a decision rather than a default.

The six measurements

This is the part you can run yourself, on any of the three, with one subject you know well. Our reference bands come from ten high-retention vertical documentaries measured frame by frame — seven English, three Turkish — and one full production manifest.

1. Duration. Reference band: 36.8 to 58.3 seconds. Nothing in the set was shorter than 36. A tool that pads to a round 60 or clips at 30 is following a template, not the story.

2. Median shot length. Reference band: 2.03 to 6.00 seconds. Under two seconds throughout usually means dead clips are being hidden behind speed. Six seconds is fine when something is developing inside the frame.

3. Cuts per minute. Reference band: 8.2 to 25.1. Both extremes held attention, so there is no correct number — what you are testing is whether the cuts land with the sentences or on a timer. Watch once with audio only, once with video only.

4. Source count. Our measured render b2e4b3eb used five separate archives for a 53.47-second piece with 14 cuts. One source is a warning: material from a single library shares a grade and a framing convention, and that uniform surface is most of what people mean by "AI slop".

5. Black frames and silence. Zero and zero, in all ten reference videos. Step through every transition. A single dark frame reads as the video ending.

6. Factual accuracy. Pick two claims from the narration and verify them. This is where fully generated pipelines fail most quietly, and it is the failure with the largest downside.

The cost question, honestly

Subscription prices for all three change often enough that printing them here would be wrong within a quarter. Check them at the source.

What almost nobody publishes is production cost, so here is ours: a median of $0.185 per finished Short across 20 ledger entries, a mean of $0.181, ranging from $0.020 to $0.414. That is what the machines cost, not what anyone charges.

The number is useful for a reason beyond curiosity. It explains why credit limits exist across this whole category, why the generous-looking plan throttles, and why no serious tool here offers an unlimited free tier. When you see a plan that appears to break that arithmetic, the constraint has usually moved somewhere less visible — queue priority, resolution, watermarking, or how quickly "unlimited" becomes rate-limited.

What each one is actually for

The useful ending is not a winner but a fit.

Choose InVideo if you need breadth — several formats, several aspect ratios, one subscription — and you are willing to do more shaping to get a specific look.

Choose Revid if throughput on short-form social is the constraint and generated visuals suit your subject.

Choose ViewMade if the video makes claims about real events and you need real footage with named sources — and accept that a tool built for vertical documentary Shorts will not be the one to make your product demo.

The failure mode is not picking the wrong one. It is picking a tool built for one of those jobs and grading it on another.

<!-- faq -->

Questions people ask

InVideo or Revid — which is better for faceless YouTube?

They differ less in quality than in starting point and visual source: InVideo is the broader general-purpose assembler from a prompt or script, Revid is oriented to fast short-form output with generated visuals. Run the same subject through both and measure duration, median shot length, cut placement against the narration, and source variety rather than comparing feature lists.

How do I compare AI video tools fairly?

Give each one the same subject you know well, then measure rather than react: duration against a 36.8–58.3 second reference band, median shot length against 2.03–6.00 seconds, whether cuts land mid-clause, how many distinct sources the footage came from, whether any black frames or silence appear, and whether two claims in the narration survive checking.

Why does ViewMade not upload videos to YouTube?

Uploading requires a much broader permission on your channel than reading it. ViewMade requests read-only access, discards the access token after the import, and hands you the file to publish yourself. Tools that schedule for you are making a different trade, which is worth making deliberately.

What does one AI Short cost to produce?

In our ledger of 20 jobs, a median of $0.185 and a mean of $0.181, ranging from $0.020 to $0.414. Subscription prices are a separate question and change too often to state usefully in an article.

Is generated footage or archive footage better?

Generated footage is stronger when the subject never existed in front of a camera and weaker when it did. The recurring failure is using a single source for everything — that uniform grade and framing is what makes many faceless channels look identical to each other.

Does this article rank the three tools?

No, and deliberately. We have not run the other two through our measurement pipeline, so scoring them would be the same unsupported comparison this category is already full of. The six measurements above are the part you can run yourself.