How to make UGC ads with AI without a new avatar vendor

This is a workflow pillar: how a performance team or a UGC agency briefs, generates, reviews, and ships ad video on the twelve keys. It is not a second cost table. Cost per approved variant lives on AI UGC ads cost. The buying guide is best AI video generator for ads. Same-prompt footage is a compare still to come, and it needs a real run first. This page has no lotgen.ai run, no ranked quality scores, and no “we tested 50 variants on lotgen” claim. Bulk numbers below are a worked model.

The short answer

How to make UGC ads with AI, as a production workflow rather than a prompt cookbook. Dated 2026-09-18.

1. Brief the take. Hook, product, billed length, aspect, audio on or off. Write the destination first (Meta Feed is below). TikTok’s own spec page is the source for a TikTok pair, not this one.
2. Pick the key before you prompt. Model and tier, on a key that will actually bill that length. Veo 3.1 is 4, 6, 8 seconds in one take. Kling vs Veo dollars for an 8-second clip are on Kling vs Veo cost.
3. Generate at the length and aspect the destination will run. A 9:16 keep that also has to run in Feed at 4:5 is a second render of the keep, not a surprise crop in Ads Manager.
4. Review the takes. Share a review link. The client sees the cut. They do not need a credit screenshot.
5. Pick an approved variant. Mount the keep. Export ships the mounted take, not “whatever rendered last.”
6. Export, then spend the aspect re-render and the paid media.

Twelve keys lotgen.ai pastes: Google, BytePlus ARK, fal, evolink, Kling, OpenAI, MiniMax, Black Forest Labs, Alibaba, ElevenLabs, xAI and Runway. Luma and Sora are models on fal or evolink, not keys. BytePlus ARK is the console that issues the Seedance key, and the name to put on the invoice.

What “UGC ads with AI” means here

On this page, a UGC ad is product footage, a talking-head, or b-roll generated on those twelve keys, then reviewed and shipped as a variant a media buyer can run. The script can still sound like a creator. The invoice is still dollars and seconds on a lab key.

That is a different job from hiring an avatar actor.

lotgen.ai is a control plane for teams who already generate with Kling, Veo, and friends and need the work to be reviewable and billable. Invite-only beta. You still pay the provider. lotgen does not resell Kling credits or Veo tokens.

On the live product (the homepage and the six studios), UGC Factory is the studio that plans that kind of lot: “The script is the ad,” a structure, a hook and a line per beat. Its card reads 20 hooks · 11 structures, video · dialogue. Marketing Studio is the sibling when you have a product page and want a shot list. After either studio plans the lot, the rest of the product is the same: shots, takes, a mounted keep, a review link, an export. None of that is a lotgen run of 50 ads. It is how the software is built.

OpenAI’s key in this product pays for GPT Image stills, not Sora. ElevenLabs is voice, not picture. Both are keys, so they stay on the list; they are not video generators. Sora 2 is a model you reach through fal or evolink.

Do not repeat vendor “AI vs human creator” savings claims as fact. If a vendor page advertises that gap, it is a vendor claim. Cite it only as such, with the human invoice in hand.

Pick the key before you prompt

The prompt is not the first line on the job. The billed length is. A 15-second brief on a model that maxes at 8 seconds is a stitch, an edit, or a different key. Write that down before anyone types a hook.

Veo 3.1 is 4, 6, 8 seconds. Google’s Veo 3.1 generate docs on 2026-09-18: video lengths 4, 6, 8 seconds; reference image-to-video only supports 8 seconds. Aspect ratios listed there: 16:9 and 9:16. 1:1 is not on that card. Audio is billed as the default Gemini row on Gemini API pricing. Gemini vs Vertex: dollar cells on this page use Gemini API, paid tier, audio default. Vertex prints a silent column that this page does not mix in. The full Standard / Fast / Lite walk is Veo 3 cost per second. “Veo 3.1” without Fast / Standard / Lite is a 4× invoice gap at 8 seconds ($0.80 vs $3.20 per take at 720p on Gemini).

Kling 3.0 can be the 15-second row. Official developer pricing on 2026-09-18: 1 Unit = $0.14 for video. Kling 3.0 720p is 0.6 Units/s silent ($0.084) and 0.9 Units/s with native audio ($0.126); 1080p is 0.8 / 1.2 Units/s ($0.112 / $0.168). Kling’s public t2v HTML did not print a duration table when checked. The 3-15s range used below is evolink’s published copy plus lotgen’s own test (2026-08-25), labeled as such. 8-second Kling vs Veo dollars: Kling vs Veo cost. Kling credits vs API Units: Kling pricing explained. A Kling website membership does not fund a Kling developer key.

15-second approved-variant math is the cost page, not this one. AI UGC ads cost owns it. The short 50-variant table later on this page is a pointer, not a rewrite of that grid.

Other length clamps that change the brief, from official pages fetched 2026-09-18:

• MiniMax Hailuo 2.3 PAYG list: 768p at 6s ($0.28) or 10s ($0.56); 1080p 6s only ($0.49). Not 15 seconds in one take. PAYG.
• Alibaba Wan 2.7 t2v International: 720P $0.10/s, 1080P $0.15/s on Model Studio pricing. Official t2v duration is an integer from 2 to 15 seconds (default 5) on the text-to-video API. Live lotgen /pricing still lists Wan 2.7 1080p as unpriced. That is the product catalog, not Alibaba list.
• evolink Kling 3.0 page copy: generate 3-15 second videos. Marketplace. Base $0.08/s; 720p audio on is $0.12/s on the day checked. Do not paste evolink’s “official Kling $0.089” line as Kling list. Kling official silent 720p is $0.084.
• Black Forest Labs FLUX 3: up to 20 seconds on the overview. Official hd $0.17/s, fhd $0.29/s, draft t2v hd $0.06/s.

Tier is part of the key. “Veo 3.1” without Fast / Standard / Lite is a 4× invoice gap at 8 seconds ($0.80 vs $3.20 per take at 720p on Gemini). “Kling” without API Units vs website credits is how “Kling is $X per video” pages disagree. Get both in writing.

Product fact (not lab marketing): US lotgen workspaces cannot use BytePlus ARK direct. The Seedance row still belongs for keys that can paste it. US readers take fal/evolink Seedance, at marketplace rates.

Same-prompt Kling vs Veo footage is a compare still to come. It needs a real run first. This page does not invent it. The live buying guide best AI video generator for ads owns “best for ads” as cost per approved variant. This page owns the steps.

Brief the take

Five fields, in this order. A prompt without them is a sketch of a different job.

1. Hook. The first seconds the destination will actually play. Not the product manifesto.
2. Product. What has to be on camera: pack, UI, a pair of hands, a talking-head. Likeness and usage rights are not a per-second row. If a named person has to be the actor, that is the avatar aside above, or a real shoot.
3. Length. The length the API will bill, which is not always the length in the media plan. Veo 3.1: 4, 6, 8. MiniMax Hailuo 2.3 at 1080p: 6. Kling 3.0: treat 15s as allowed only with the duration label above. If the brief is “15-second Reels” and the key maxes at 8, budget a stitch (two clips, then you cut) or change the key. Stitching is an edit, not a list price.
4. Aspect. Generate at the ratio the placement will run, or plan the second render of the keep.
5. Audio. Native soundtrack on or off, and whether a voice line is a separate ElevenLabs key. Veo 3.1 on Gemini bills audio as the default row. Kling 3.0 has a silent row and a native-audio row (720p: 0.6 vs 0.9 Units/s). Do not brief a silent Kling cell against an always-on Veo cell and call it a fair ad-stack compare.

Meta Feed specs (fetched 2026-09-18). Facebook Feed video: file MP4, MOV, or GIF. Recommended ratio 4:5 (1440 × 1800). Tooltip on that ratio: ratios from 16:9 to 9:16 are supported but may be masked to 4:5; on desktop Facebook Feed the video may show at 1:1, which can add bars. Duration 1 second to 241 minutes, so 15 seconds and 8 seconds are both inside that window. Sound and captions: optional, but recommended. Max file 4 GB.

TikTok’s own ad specs were not reachable on 2026-09-18. This page does not treat 9:16 / 15s as a TikTok-official pair; read TikTok’s own ads-spec page before you brief a Reels-style cell.

A 9:16 keep that also has to run in Feed at 4:5 is one extra take of the keep, unless the crop fails and you regenerate. Full re-generation at the second aspect is the conservative model. Pick it in writing.

On-pack legal copy, claims, or a logo that must be pixel-true is a different QA bar. Provider docs checked for these pages do not publish a “text rendering SLA.” A keep-rate model assumes you can throw a take.

Generate in bulk without a new vendor

You do not need a new avatar vendor to run fifty prompts. You need the twelve keys you already pay for, a billed length that matches the brief, and a place to keep the takes.

This is a worked model, not a lotgen.ai run. No prompts were submitted. No takes were kept.

approved_variants = 50          # one prompt family each, one aspect
attempts          = 3           # modeling assumption, not a measured keep rate
takes_to_run      = 50 × 3      # 150 renders

A studio-shaped version of the same model, using UGC Factory’s published hook and structure counts only as capacity copy, not as a test we ran:

# The studio's card: 20 hooks, 11 structures.
# You will not ship 220 ads. You will pick a board.
hooks_you_keep      = 10
structures_you_keep = 5
board               = 10 × 5 = 50 variants
takes_to_run        = 50 × 3 = 150

Retries are the first multiplier. A missed label, a bad hand, a talking-head that does not hold: you pay for the miss and the keep. Public provider pages checked for this page do not publish a keep rate. 3× is a slider you set. Change it. Do not cite an “industry keep rate.”

Aspect re-renders are the second multiplier. They apply to the keep, not to the whole 3-attempt stack, if you crop-then-check:

second_aspect_of_the_keep = 50 × 1 take

If you regenerate from scratch in the second aspect, that line becomes another 150 takes. Say which one you quoted.

fal and evolink are their own checkout. Matching Kling’s 720p audio row on fal is a coincidence on the day we last checked that marketplace, not a promise that 1080p or 4K match. Luma and Sora sit in those catalogs. Nobody pastes a Luma key or a Sora key here.

Re-render one shot, not the lot. The Academy lesson re-rendering one shot: a lot is a list; fixing one note costs one shot; the take you replace stays, because you paid for it. That is the bulk-testing mechanic. It is not a claim that we ran 50 variants.

Failed generations: Kling’s official developer page says failed video generations do not deduct Units. Gemini Veo: charged only if the video is successfully generated. MiniMax PAYG copy: failures / security-review videos do not deduct. Other labs differ. Quote the successful take.

Review like a client

A review room exists so a client or a brand manager can sign off the cut without logging into the studio and without seeing what you paid.

Live product copy, checked 2026-09-18:

• Homepage: “A review room with no login. They see the cut, never your costs, providers or prompts.”
• Pricing, Solo: “Review links for clients, with no login for them.” Team: “Winners, review rounds and approvals with a name on them.”
• Academy lesson review links: a review link is a public URL for one lot. No login, no account, no app. Anyone who holds the token can watch and cast a verdict, which is why you give the link an expiry. What they see: each shot’s mounted take, the shot name, the prompt, the narration. What they are never shown: cost, provider, model, the editor’s notes. Reviewers see each other’s names and verdicts. Written comments stay between them and you.

Standing, as that lesson defines it:

Only a reviewer’s latest verdict counts. One outstanding change request holds a shot. Unreviewed does not block, and it also is not done. “Four of six approved” is not a lot ready to deliver.

Mount the keep before you share. The Academy lesson what a lot is: the first successful take of a shot mounts itself. That is a one-take convenience, and it means the take on show is the first one that finished, not the one you would have chosen. When you render several variants of a shot that has no winner yet, the one that comes back first becomes the shot’s answer. Overrule it. Export and the review link both show the mounted take. Do not send an unfinished lot.

What the product marks, and what it does not. There is no “Reviewed in LOTGEN” badge on a delivery. What there is:

• The caption on the homepage’s clip: Rendered in LOTGEN (a caption on the marketing page, not a review-room footer).
• Every public footer: “Every export carries the EU AI Act Article 50 mark.”
• Academy lesson the export pack: the delivery is the mounted take of every shot, files named after the shots, a manifest (aspect / resolution / audio requested vs delivered, provenance), and a synthetic-content declaration. It is a declaration, not a signed C2PA manifest. The lesson says a signed C2PA guarantee would need a certificate the workspace has not been given.

Approve, then export. The export pack is supposed to ship the same take the client approved. That is the whole point of mounting.

Cost of a 50-variant test

One short table. The full 15-second approved-variant grid, aspect re-renders, and the client-round model live on AI UGC ads cost. This is not that page.

This is a worked model, not a lotgen.ai run.

billed_seconds     = clamp(requested_seconds, provider min, provider max)
cost_per_take      = list_rate × billed_seconds
cost_per_approved  = cost_per_take × attempts     # one aspect, before review
batch_total        = cost_per_approved × 50

Assumptions you can change: 50 approved variants, one prompt family each, audio on / default, one aspect, 3 attempts. Quality is not in the table. The Veo row is Gemini, audio default; Vertex’s silent column is not mixed in.

Rounded to the nearest dollar in the batch column so a media plan can scan it. Exact cents: Kling 720p audio 3× × 50 = $283.50; Veo Fast 720p 3× × 50 = $120.00; MiniMax 768p 10s 3× × 50 = $84.00; Wan 720P 3× × 50 = $225.00.

Worked Kling 3.0 API, 720p, native audio, 15s, 50 approved, 3 attempts

0.9 Units/s × $0.14 = $0.126/s
$0.126 × 15s = $1.89 per take
$1.89 × 3 = $5.67 per approved variant
$5.67 × 50 = $283.50

Worked Veo 3.1 Fast 720p, 50 approved, 3 attempts, honest 8s cell (Gemini audio default)

$0.10/s × 8s = $0.80 per take
$0.80 × 3 = $2.40 per approved 8s clip
$2.40 × 50 = $120

That $120 is not a 15-second variant budget. It is fifty approved 8-second clips. A stitched ~15s model (two 8s takes per attempt, then you cut) is $240 at Fast 720p before aspects and review. Edit time is extra. Not a lotgen run. Not a claim that the join will look like one shot.

Live 8-second Kling vs Veo walkthrough: Kling vs Veo cost. Veo Standard / Fast / Lite: Veo 3 cost per second. The cost pillar: AI video generation cost. The calculator link below opens with the agency preset: 15 seconds, 1080p, audio on, 50 shots, 3 attempts, the client toggle on. Pick the tile that fits, Agency or studio or Brand or in-house team. Veo rows will clamp to 8 billed seconds. The 3-attempt slider stays a model you set.

How to quote before you spend

lotgen.ai is a control plane, not the point of this page. One place to generate takes across the twelve keys, see a real-dollar quote before spending, compare outputs side by side, share a review link, and drive it via MCP. Invite-only beta. You still pay the provider. Seconds bill to your keys, at list. Seats are a separate line.

Public seat copy on lotgen.ai/pricing on 2026-09-18: $99 Solo, Team $90 a seat from 3, $79 from 6. Two seats is not a plan.

A quote that holds up in a monthly recap is dollars, seconds, model, tier, attempts, on the keys you will actually paste.

1. Brief the variant: billed length the destination will run, aspect (4:5 Feed vs 9:16), audio on or off, model and tier.
2. Quote the take in dollars on those keys, before the job runs. The Academy lesson what a render costs: two calls spend money (render a shot, render a lot, plus paid fixes). Planning, attaching references, choosing a model, and pricing the campaign are free. A shot written for seven seconds on a model that only makes four, six, or eight is quoted for eight, because eight is what you will get.
3. Generate across more than one key when the brief allows it. Same prompt, two invoices, pick a keep. This page has no side-by-side render to show you. The same-prompt compare will, when a lotgen run exists.
4. Compare the takes with the quote still attached.
5. Share a review link so a client or brand manager signs off without a credit screenshot. Mount the keep first.
6. Approve, export the mounted take, then spend the aspect re-render and the paid media.

Request an invitationSeats and pricingRequest a team seat on the beta.

Brand / in-house teams can use the same form: request an invitation. Seat and plan detail: lotgen.ai/pricing. This does not create an account and it charges nothing.

This page is not a live lotgen.ai invoice and not a 50-variant test. Replace the model with real quotes before you treat a cell as a forecast.

Prices checked on 2026-09-18: how to re-check yourself

Refresh this page on each major model release, minimum quarterly. Prefer the live cost siblings over copying their grids here.

1. Kling Units and per-second rows: kling.ai/dev/pricing. Confirm 1 Unit = $0.14 for video, then the Kling 3.0 720p/1080p cells with and without native audio. Re-read duration on Kling’s own video API docs; this page used evolink’s 3-15s copy plus lotgen’s own test.
2. Veo 3.1 Gemini per-second rows: Gemini API pricing, section “Veo 3.1”. Confirm Standard / Fast / Lite. Google’s page was last modified 2026-09-16 when checked. Dollar cells here use Gemini audio default, not Vertex silent. Lengths and aspect: Veo 3.1 generate (4, 6, 8 seconds; 16:9 and 9:16).
3. MiniMax PAYG dollars: pricing-paygo. Use PAYG as list. Hailuo 2.3 still not 15s in one take.
4. Alibaba: Model Studio pricing and text-to-video API. International wan2.7-t2v is $0.10/s 720P and $0.15/s 1080P. Duration 2 to 15s. Do not paste evolink Wan as Alibaba official.
5. evolink Kling (marketplace only): evolink.ai/kling-3. Ignore their “official Kling $0.089” line unless Kling’s own page has moved to that number.
6. Black Forest Labs: docs.bfl.ml/quick_start/pricing and FLUX 3 overview for the 20-second cap.
7. Meta ad specs: Ads Guide, Facebook Feed video. TikTok’s own ads-spec page was not reachable when this was checked; read it there before you plan a Reels cell.
8. Live cost siblings, rather than copying their grids: AI UGC ads cost, Kling vs Veo cost, Veo 3 cost per second, AI video generation cost, best AI video generator for ads, AI video cost calculator.

Recompute: $/s × billed seconds × attempts × approved variants. If the billed seconds are not a length that provider accepts, stop and check the model id.

Related on this site: AI UGC ads cost · Kling vs Veo cost · Kling pricing explained · AI video generation cost · Veo 3 cost per second · best AI video generator for ads · AI video cost calculator · how a render is quoted before it runs · review links · the six studios · seats and pricing · request an invitation

The MCP how-to: generate AI video from Claude Code.

Request an invitationSeats and pricingA dollar quote on your own keys before a 50-variant board runs.

FAQ