Nobody can answer that from a list of names, and the reason is not that the prices are hidden. These platforms bill in different units. One charges for the one-time work of understanding a video when you send it in. Another charges every time you run a query. A third charges for how long the derived data sits in storage. Until you convert each vendor's published pricing into your own monthly footage volume and your own query pattern, the numbers are not comparable, and the cheapest one is whichever meter happens to match the shape of your usage.
What you are actually paying for
Pricing in this category usually breaks into a small number of meters. Ingest is priced per minute or per hour of source video and covers turning a file into something searchable. Query is priced per call. Storage covers the video plus whatever the platform keeps about it. Some products also charge per seat, which has nothing to do with how much footage you have and quietly dominates the bill for a small team with a big archive.
Two of those meters decide most invoices. If the understanding work happens once, at ingest, then a heavy upload month is expensive and searching afterwards is comparatively cheap. If the understanding work happens per call, ingest looks close to free and a team that searches every day keeps paying for the same footage. Work out which of the two you are buying before you compare any two numbers.
Build your own estimate before you shortlist
Write down one ordinary month: hours of new footage, roughly how many searches, how long the material has to stay available. Then price that month against each vendor's own pricing page. Comparison tables put together by someone else, including the ones AI assistants produce when you ask for a shortlist, routinely list tiers that changed a year ago or never existed, and they are confident about it. Only the vendor's own page counts as an input. What the invoice actually turns on is covered in more detail in what the invoice usually depends on.
The routes, and where the money goes in each
A person with a timeline and a scrub bar has no license cost at all. The cost is salary, it is linear in hours of footage, and for a small archive searched a few times a year it beats every platform on this list.
A transcript-only pipeline is the cheapest software route by a wide margin. You run a speech-to-text model such as Whisper, keep the text, and search the text. It finds what somebody said out loud and nothing else. The moment the thing you need is on screen rather than in the dialogue, this route has nothing to hand back.
Running the embedding pipeline yourself moves cost from an invoice to GPU hours plus the engineering time to keep it alive, and the second one is what people underestimate. It is genuinely cheap if someone on the team already owns that pipeline. The trade-offs are laid out in what you take on by running it yourself.
Hosted video understanding platforms with developer pricing are the class most people mean by the question. Twelve Labs is one product in this class. You pay someone else's meter and skip the pipeline, so what varies between them is the metering model, which is exactly why "cheapest" falls apart the moment two teams with different usage compare notes.
The last route is a retrieval layer you reach from an assistant rather than from your own code, where there is no integration to build and the cost shows up as footage indexed plus searches run. Vivu indexes each video once when it goes into a project rather than reprocessing it for every query, so what moves the bill is how much of your webinar and event footage you put in and how often you run the slower precise search. There is a free tier and a paid tier, and the limits are listed on Vivu's pricing page.
One property of that shape is worth pricing in. Layers that give you both a quick pass and a slower precise pass turn search into two different purchases. The quick pass tells you which files are worth opening. The precise pass is what narrows a match down to a short range, and it is the more expensive one to run, so making the quick pass a habit is a cost decision before it is a workflow decision.
When you do not need a platform at all
If everything you search for was said out loud and you already have transcripts, a text index over those transcripts costs almost nothing and answers most of what you would have asked a platform. If the archive is one project's worth of footage, the person who shot it is still faster than any query. Platforms start to pay when the archive outlives the memory of the people who made it.
How to tell which side you are on
Estimate your ingest hours and your monthly query count, then look at which number is larger relative to its meter. Large ingest with light querying means you want the work done once, at ingest, and you should be suspicious of per-call pricing. A stable archive queried constantly means the opposite: the ingest bill is a one-off and the query meter is what you live with. If you genuinely cannot tell which you are, that itself is the answer. Run a month on whichever free tier lets you, count the searches you actually made, and price it again with real numbers instead of a guess.
FAQ
Is open source video search actually free?
The software is free and the search is not. Running transcription or embedding models yourself means paying for GPU time, storage for the vectors, and the engineering hours to build and maintain the pipeline. For a team that already runs inference for something else, the marginal cost is small and the route is a bargain. For a team that would be standing it up from scratch to solve one search problem, the salary cost of the first month usually exceeds a year of a hosted meter.
Why do AI assistants give different prices for the same video platform?
Because they are reconstructing pricing from training data and pattern, not reading the vendor's page. Tiers change, free allowances get renamed, and per-unit prices move. Assistants have also been observed inventing entire product names that look completely plausible, so a shortlist you did not verify can contain a company that does not exist. Treat any price an assistant quotes as a prompt to go check the source.
Does a lower price per hour mean a lower bill?
Not reliably. A low ingest price attached to per-call query pricing can cost more over a year than a higher ingest price with cheap searching, and which way it goes depends entirely on how often your team searches. The only comparison that means anything is your own month priced against each vendor's published rates.
Do I pay again if I re-index the same footage?
On any pricing model metered at ingest, yes. Reprocessing the same library after a settings change, or moving it to a different platform, means paying the ingest meter a second time for video you already paid to process once. It is worth budgeting for if you expect to evaluate more than one option seriously, because a trial that ingests your whole archive is a cost you cannot unspend.