Commercial advertising is where AI video either proves itself or gets laughed out of the room. A brand paying for a 30-second spot doesn’t care that your clip was generated in four minutes. They care that it holds up on a television screen, matches their brand book, clears legal, and delivers on the media plan. If you want to sell AI video to brands or agencies, you need to speak their language: specs, approvals, revisions, deliverables. That’s it. That’s the whole game.
This guide covers what commercial work actually demands — resolution and bitrate minimums, brand compliance, approval workflows, revision discipline, working with media buyers, and format-by-format breakdowns of the 6-second bumper, the 15-second cutdown, and the 30-second hero spot. Pricing and platform specs shift constantly, so treat numbers here as working baselines — and verify against the current spec sheets before you quote.
What brands expect (and why AI video gets judged harder)
Here’s the uncomfortable truth: AI-generated footage gets scrutinized more than shot footage. A slightly soft background in a live-action commercial reads as depth of field. The same softness in an AI clip reads as “this looks fake.” Brands know their audiences are alert to AI tells — warped hands, melting text, faces that drift between cuts — and they will reject work that a casual viewer made with a free tool would accept. Different audiences, different mercy.
So your quality bar for commercial work sits one level above your quality bar for social content. Every artifact you could ignore on TikTok becomes a revision note on a brand job. Plan for it. Budget for it. Don’t let it surprise you.
Technical minimums: the specs that matter
Before anything creative, lock the technical baseline. These are the minimums most brands and broadcasters expect in 2026:
Resolution
- 1080p (1920×1080) is the floor for anything going to broadcast, YouTube pre-roll, or connected TV. No exceptions.
- 4K (3840×2160) is increasingly the ask for hero spots, especially from brands that repurpose one master across broadcast and digital.
- Most AI generators output at 720p or 1080p natively. If your tool tops out at 720p, you need an upscaling step in your pipeline — Topaz Video AI and similar tools are standard practice, not cheating. Budget the render time.
- Always confirm whether the client wants the master in 4K. Upscaling after approval is fine; delivering a 1080p master when the media plan assumed 4K is a credibility hit.
Frame rate
- 24fps for cinematic spots. 25fps for broadcast in PAL regions. 30fps is common for US digital and broadcast.
- Match the frame rate to the delivery spec, not to whatever your generator defaults to. Converting frame rates in post is routine, but plan for it — 24-to-30 conversion done badly creates judder that clients notice.
Bitrate and codec
- For broadcast delivery, stations and networks publish exact specs — often ProRes 422 or XDCAM HD at 50 Mbps for HD, higher for 4K. Get the spec sheet before you start, not after you finish.
- For digital (YouTube, Meta, TikTok), upload the highest quality master you have and let the platform compress. A good rule: export H.264 or H.265 at 20–50 Mbps for 1080p, 50–100 Mbps for 4K.
- Never deliver the raw generator output as a final file. AI tools compress aggressively; your post pipeline (edit, grade, upscale, export) is what gets you to delivery quality.
Color
- Deliver in Rec. 709 for broadcast and standard digital. If the brand works in HDR (increasingly common for premium spots), confirm the HDR spec — HDR10, Dolby Vision — and whether your pipeline can actually produce it.
- AI generators do not output calibrated color. Your grade is doing real work here: matching shots to each other and to the brand’s look.
Brand guideline compliance
Every brand of consequence has a brand book — colors, fonts, logo usage, tone of voice, things you may never show. And AI video introduces specific compliance risks that traditional production doesn’t:
- Logo rendering: AI models are notoriously bad at rendering exact logos and text. Never generate a logo inside the video model. Composite real logo assets in post, exactly as the brand book specifies — clear space, minimum size, approved lockups only.
- Brand colors: Do not trust the generator to hit a Pantone. Generate neutral, then match brand colors precisely in the grade using the brand’s hex or Pantone values.
- Product accuracy: If the spot shows a product — a bottle, a phone, a car — the product must be accurate. AI will happily invent a label, change a button layout, or redesign packaging. Use image-to-video with the actual product photography as reference, or composite the real product into generated backgrounds.
- People and representation: Brand books increasingly specify representation guidelines. AI casting is your responsibility — the faces you generate are the faces the brand is associated with. Get sign-off on casting direction early.
- Tone: A luxury brand and a fast-food brand do not move the same way. Camera motion, pacing, and energy should be argued from the brand book, not from what the model does well.
Get the brand book before you generate a single frame. Read it. Actually read it. Reading it after the first cut is how you end up regenerating everything.
Approval workflows that survive contact with clients
AI video’s speed is a trap in commercial work. Because you can generate twenty variations overnight, clients expect twenty variations overnight — and then want to see twenty more. Without a structured approval process, you will drown in revisions. It’s not a possibility. It’s a certainty.
Stage your approvals
- Concept and boards: Storyboards or style frames first. Get sign-off on the idea before any generation. This is the cheapest stage to change direction.
- Animatic or previz: Rough motion — even crude generations — to approve pacing, camera moves, and structure. Label it clearly as non-final quality.
- Hero frames: One polished still per key shot, approved for look, casting, product, and brand compliance before you generate full motion.
- Full-motion draft: The complete spot at final timing, with temp sound. This is where motion-specific notes land.
- Final master: Graded, mixed, with final logos and legal lines. Approvals here should be limited to technical QC — if creative notes arrive at this stage, your earlier stages failed.
Rules that protect you
- One consolidated round of notes per stage. Five stakeholders sending separate emails is how you get contradictory direction. Insist on a single compiled notes document.
- Define what a “revision” is in the contract. A color tweak is a revision. “Let’s try a completely different concept” is a new project. Write this down before work starts.
- Version everything. v1, v2, v3 with dates. When a client says “go back to the version from Tuesday,” you need to know exactly which file that is.
- Watermark drafts. Timecoded watermarks on every draft. It prevents unfinished work from leaking and reminds everyone what stage they are looking at.
Revisions at scale
Now, AI video does change the economics of revisions — regenerating a shot costs minutes and credits, not a reshoot day. That’s a real advantage. But only if you manage it:
- Save every prompt, seed, and setting. A revision request like “same shot but raining” is trivial if you have the original generation parameters. It is a full rebuild if you don’t. Treat prompts as production assets, filed per shot.
- Batch your regenerations. Collect all revision notes, then run one generation session. Firing off single generations per note wastes time and makes version tracking chaotic.
- Know when to stop generating and start compositing. Small fixes — a logo swap, a color shift, removing an artifact — are faster in After Effects or Resolve than in the generator. Regenerate for motion and performance changes; fix in post for everything else.
- Charge for revision rounds beyond the agreed number. AI makes revisions cheap for you, which makes clients request more of them. Your contract should include a set number of revision rounds, with additional rounds billed. Cheap revisions still cost your time.
Working with media buyers
Media buyers plan where the spot runs — and each placement has its own spec. The sooner you talk to the buyer (or get their spec sheet), the fewer surprises at delivery. So talk early:
- Get the full placement list early: broadcast, YouTube pre-roll (skippable vs. non-skippable), Meta feed, TikTok, connected TV, digital out-of-home. Each has different duration, aspect ratio, and technical requirements.
- Aspect ratio variants are real deliverables. A 30-second 16:9 hero spot usually needs 1:1 and 9:16 cutdowns. AI generation is actually an advantage here — generating or reframing for vertical is cheaper than reshooting. But each variant needs its own QC pass; do not assume a center-crop is acceptable.
- Closed captions and legal lines: Broadcast spots need captions; many digital placements do too. Legal disclaimers (“terms apply,” contest rules) must be on screen for the required duration at the required size. AI cannot generate reliable on-screen text — add all text in post.
- Delivery destinations: Ask where files go — Extreme Reach, Comcast AdDelivery, YouTube directly, the client’s DAM. Each has upload specs and lead times. Missing a traffic deadline because you exported the wrong codec is an amateur mistake.
Format breakdowns: 6s, 15s, 30s
The 6-second bumper
Six seconds is one idea, stated once. No setup, no story arc — a single striking image or motion with the brand present throughout. YouTube bumpers are unskippable, which means the audience can’t escape. Respect that with clarity rather than noise.
- Structure: Brand/product visible within the first second. One motion or transformation. End on logo or product.
- AI considerations: Six seconds is within a single generation for most current models, which means maximum quality per second — spend your credits here on the highest settings. Character consistency is rarely an issue at this length.
- Common failure: Trying to fit a 15-second idea into 6 seconds. If the board has three beats, it is not a bumper.
The 15-second cutdown
Fifteen seconds allows a simple arc: hook, turn, payoff. Most 15s spots are cut down from a 30s hero — but AI workflows let you generate the 15s as its own edit rather than a literal cut. Often better, honestly, because pacing designed for 30 seconds feels rushed when compressed.
- Structure: Hook in the first 3 seconds (skippable pre-roll demands it), one clear message, brand payoff at the end.
- AI considerations: Two to three shots. This is where shot-to-shot consistency starts to matter — lock your character and style references before generating. Budget for regeneration: the turn (middle beat) usually needs the most iterations.
- Common failure: The message gets lost. Read the script out loud at 15-second pace; if you cannot say the key line comfortably, the spot is overstuffed.
The 30-second hero
Thirty seconds is a complete miniature story: setup, development, payoff, brand button. This is also the format where AI video’s weaknesses are most exposed. Five to eight shots means five to eight chances for consistency breaks — and the audience has time to notice. They will notice.
- Structure: Setup (0–8s), development or demonstration (8–20s), payoff and brand (20–30s). The brand should appear or be implied throughout, not just slapped on the end.
- AI considerations: This is a pipeline job, not a generation job. Storyboard every shot. Generate with locked references. Edit for rhythm first, then fix consistency issues in post. Expect the full pipeline — generation, edit, grade, sound, delivery — to take days, not hours, if you want broadcast quality.
- Common failure: Inconsistent product or character across shots. Assign one person (even if that person is you) to watch every cut specifically for continuity, separate from the creative review.
The delivery checklist
Before any commercial master leaves your hands, run this:
- Resolution, frame rate, and codec match the delivery spec sheet.
- All logos are real composited assets, not generated approximations.
- Brand colors verified against the brand book values.
- Product depictions checked against real product references.
- On-screen text spelled correctly, held for the required duration.
- Legal lines and disclaimers present and legible.
- Captions file prepared where required.
- Audio levels meet spec (broadcast loudness standards are strict — typically -24 LUFS for US broadcast).
- All aspect ratio variants rendered and QC’d individually.
- File naming follows the traffic/delivery convention.
- Client sign-off documented in writing on the final master.
Here’s the thing about commercial AI video work: it’s not about the novelty of generation. It’s about everything around the generation. The creators winning brand work in 2026 aren’t the ones with the flashiest clips — they’re the ones whose delivery process looks exactly like a traditional production company’s, with AI as the engine inside. Build that process, and the tool becomes irrelevant. Skip it, and no model will save you.
For the cost side of commercial projects — budgeting credits, subscriptions, and post-production time into client quotes — see our AI video cost and pricing guide. And if you’re evaluating which generator to standardize on for client work, start with our comparison of the top AI video generators.