Creators and ecommerce buyers keep arguing about the wrong number. One tab advertises a video as low as a few cents. Another tab advertises a richer route for more. The invoice line looks decisive until you notice nobody wrote what counts as a keep. That is why I tested Viddo AI as a keep-rate problem, not as a shopping list of model logos.
The test protocol is narrow on purpose. Freeze one brief. Run a cheap hunting bank and a reference keep bank. Mark keeps or kills after each take. Stop when one clip is publishable or when the hour ends. If you cannot stop, the advertised floors are just badges. The mid-section numbers only matter after that protocol holds.
What Per Video Floors Actually Buy You
A floor is a starting price for a route, not a promise that every take will ship. On Viddo AI, public floors include Veo 3.1 Lite as low as $0.0575 per video and Seedance 2.0 as low as $0.345 per video. Credits still move with model, length, and related options. Free outputs can carry a watermark. Paid plans remove it and allow commercial use under published terms. Annual billing is advertised with a 50% OFF path versus monthly. Those are the money rules. They are not a creative brief.
I also separate exploring money from publishing money. Explore money can miss. Publish money should buy a file that can leave without a watermark argument. Teams that mix those budgets treat every almost as a maybe and then wonder why the week had no ships.
Buyers mix the two. They see $0.0575 and imagine a week of finished ads. They see $0.345 and imagine cinema. Neither number says what happens when the face resets, the logo edge wobbles, or the vertical crop cuts the product name. The floor only becomes useful after you define a keep.
I define a keep as a clip you would put in the client folder without renaming it “almost.” Soft almost kills. Pretty demos that cannot survive the real channel are killed. If your team cannot say kill out loud, the floor is decorated.
Why Unit Price Lies Without A Keep Rate
Unit price answers “what does one generation cost.” Keep rate answers “how many generations does one publishable clip actually burn.” A cheap route with a 1-in-8 keep rate can cost more than a richer route with a 1-in-2 keep rate. That math is boring. It is also how weekly desks go over budget while staring at a “cheap” badge.
| Route | Public floor | Best use | Kill signal |
| Veo 3.1 Lite | as low as $0.0575 | Angle hunting | Face or product look resets every take |
| Seedance 2.0 | as low as $0.345 | Reference-heavy keeps | You ignore uploads and prompt vibes only |
| Any free path | watermark risk | Private drafts | Watermark still on a client file |
Read the table as routing, not loyalty. The cheap floor is for finding the move. The richer floor is for locking a look when references matter. Mixing those jobs is how teams burn credits and still argue about taste.
Write Keep Rules Before You Open Model Lists
Write three lines on a sticky note: channel, one required readable detail, and the kill list. Example: “9:16 SKU hero, logo edge sharp, kill face reset.” That note is the brief. Model shopping without it is entertainment.
Also decide watermark policy before Generate. If the file must leave for a client, plan a paid path so free watermarks never enter the approval thread. A keep with a watermark is not a keep.
How To Run A One Hour Keep Rate Test
Pick one frozen brief. Do not change the product, the channel, or the required detail mid-test. Run two short banks only: a cheap hunting bank and a reference keep bank. Score keep or kill after each take. Stop when you have one publishable keep or when the hour ends.
Bank One Hunt Angles On The Cheap Floor
Open the create panel on Viddo AI. Choose a lower advertised route such as Veo 3.1 Lite. Set Video Ratio for the real channel, draft at 720p, and use the length control including the 15s option when the beat needs room. Paste one concrete motion sentence. Generate. Change only the camera move on the next take. Do not rewrite the whole brand story between clicks.
If the prompt is thin, use Generate With AI to expand keywords once. Translate if English prompting behaves better for that path. Leave Web Search off unless the brief depends on a live fact. Mark each take, keep or kill in a notes app. Hunting ends when you have a move worth keeping, not when you feel inspired.
Bank Two Lock References For The Keep
Switch to a reference-friendly route such as Seedance 2.0 when the brief needs mixed inputs. Upload Media / Reference shows image(0/9), video(0/3), and audio(0/3). Put the approved stills in. Lock Seed when a take is close so style and character appearance stay steadier on the next pass. Raise resolution to 1080p only for the keep candidate.
Score the keep against the sticky note, not against a demo reel. If the required detail fails, kill it. If it passes, download and stop. That stop is the whole point of an ai video generator price test. Without the stop, floors become a slot machine.
Count Keeps Not Pretty Almost Takes
At the end of the hour, write two numbers: takes burned, keeps saved. If the cheap bank produced zero keeps and the reference bank produced one, the richer floor paid for itself on that brief. If both banks produced zero, the brief is wrong — soft source still, vague motion line, or a channel crop the product cannot survive. Do not buy more credits to paper over a broken brief.
Write the keep rate on the sticky note next to the kill list. A team that cannot report keep rate will keep buying logos. A team that can report keep rate will change briefly before it changes the vendor. That habit is the actual upgrade.
When Price Floors Mislead Weekly Buyers
Floors mislead when you treat them as a monthly content plan. They also mislead when you compare them across unlike jobs — a five-second logo sting is not a fifteen-second spokesperson cut. Compare floors only inside one frozen brief with one kill list.
They mislead again when free watermarked drafts sneak into client reviews. The review then becomes a watermark debate instead of a motion debate. Paid unlocks exist for that reason. Use them when the file must leave the building.
Keep Rate Beats Demo Reels For Buyers
This test fits teams that ship weekly and need an honest read on cheap versus reference routes. It is a weak fit if you only need one occasional clip and already trust a single specialty path.
Write the kill list. Run one hour. Count keeps. Route Lite for hunting and Seedance 2.0 when references must hold. Viddo AI is a restrained buy when those rules make the public floors mean something besides a badge.
Write and Win: Participate in Creative writing Contest & International Essay Contest and win fabulous prizes.
About the Author
