Correction, 2026-10-02: this page named Wan 3.0 at 480p (24 credits for 5 seconds) next to Pruna p-video Draft (4) as the cheapest route for volume and called it "the cheap tier". On 8frame, Creatify Boreal (7 credits for 5 seconds at 720p), Pruna p-video (14) and Hailuo H3 Max Turbo at 480p (17) also cost less than 24; they are now listed, and the page says the Wan clips were 5 seconds where the Veo clips they replaced were 8. It ranked models by quality ("b-roll workhorse", "cutaways that carry weight", "the balance of cost and look", "worth a premium model") without having measured them; models are now listed by price and job. It did not say that Kling v3 Standard needs a start image on 8frame, or that the Veo tool opens on Veo 3.1 Fast (168 credits), not Lite. It stated audience claims as fact (generated presenters "read as filler within seconds", fully generated long-form "fails", b-roll is "the largest category by volume", a forty-clip library "covers most of what a channel needs", a shot list "takes ten minutes"); we had measured none of them, so they are now advice or gone. The disclosure answer cited a "convention among creators" and did not say what YouTube requires; it now quotes YouTube's help page, and YouTube's own format rules (16:9 player, Shorts square or vertical up to 3 minutes) are added. "Seedream 5" is the model the picker calls Seedream 5 Lite.
Correction, 2026-09-30: we said one video a week fits the $19 Starter plan and two a week fit the $49 Creator plan with room. At about 456 credits a video, one a week needs roughly 1,820 credits a month and two a week about 3,650, more than those plans carry. Fixed below.
There is no single best AI video generator for long-form YouTube; there is a price per job. We don't recommend generating the whole video. What generation does fit on a long-form channel is the connective tissue: b-roll behind a voiceover, an intro, cutaways that would otherwise be stock footage, and the thumbnail. On 8frame, an 8-second Veo 3.1 Lite clip is 54 credits with sound or 33 silent (the Veo tool opens on Veo 3.1 Fast at 168, so switch it), the lowest-priced video model is Pruna p-video Draft at 4 credits for 5 seconds at 720p, and Kling v3 Standard is 86 credits for 5 seconds with sound (57 silent) from a start image. The video's worth of clips in the table below comes to about 456 credits, $4.56 at pack rate.
TL;DR
- B-roll on Veo: Veo 3.1 Lite, 54 credits for 8 seconds with sound or 33 silent, 720p. The Veo tool opens on Fast (168), so switch to Lite.
- Lowest prices per 5-second clip: Pruna p-video Draft 4 (720p), Creatify Boreal 7 (720p, sound always on), Pruna p-video 14 (720p) or p-video 2 Pro (Speed) 14 (480p), Hailuo H3 Max Turbo 17 (480p). Wan 3.0 at 480p is 24.
- Cutaways from a still: Kling v3 Standard, 86 credits for 5 seconds with sound, 57 silent, 720p. It needs a start image.
- Long single shots: Wan 3.0 runs up to 30 seconds in one pass, 142 credits at 480p.
- Thumbnails: Nano Banana Pro at 21 credits or Seedream 5 Lite at 6.
- YouTube's formats: the standard aspect ratio on a computer is 16:9; Shorts are square or vertical and up to 3 minutes.
- Disclosure: YouTube requires it for realistic AI content, such as AI-generated footage of a real place. A thumbnail made with AI doesn't need it.
Why long-form is different from Shorts
YouTube treats an upload as a Short when it is square or vertical and up to 3 minutes long. Long-form plays in a player whose standard aspect ratio on a computer is 16:9. Shorts is a format AI generation fits easily: a few seconds, one idea. Long-form asks the viewer for minutes, and on most channels they come for a person, a voice and an argument. We have not measured retention on generated long-form, so take this as a working rule: keep the person and the script, and generate the parts that would otherwise be stock footage or a static screen.
So the question is not "which model makes YouTube videos" but "which model makes the parts of my video that are currently stock footage or a static screen." That is a real and solvable problem, and it is where the budget goes.
Generate at 16:9. YouTube's help page gives 16:9 as the standard aspect ratio on a computer and advises against adding padding or black bars to the video. On 8frame, the Veo, Kling and Wan nodes open at 9:16, so switch them to 16:9 before generating long-form b-roll. Veo 3.1 offers 16:9 and 9:16 only. Kling v3 takes the shape of its start image, so give it a 16:9 still.
What to generate
B-roll behind narration. It sits under a voiceover for a few seconds and needs to be relevant and unobtrusive rather than beautiful, so the price per clip matters most here. This is where the lower-priced models are worth a test.
Intros. Ten to fifteen seconds, made once, reused for months. Because one intro runs on every video, its cost per use is small, which makes it the place to spend on a 1080p tier such as Veo 3.1 Fast (168 credits for 8 seconds with sound) or Standard (448). See how to make a YouTube intro with AI.
Cutaways that make a point. When you say a thing and the picture should show it. The viewer is looking at these, so draft two or three takes and keep the one that shows the point. Kling v3 Standard on 8frame works from a start image, which lets you anchor the cutaway to a specific frame or product shot.
Establishing and transition shots. Between chapters, to reset attention. Wan 3.0 can run a single shot up to 30 seconds.
Thumbnails. The image people see before they decide to click.
What not to generate
Your face and your delivery. The channel is the person. Keep the presenter real. Making a real person appear to say or do something they didn't is one of the cases YouTube says must be disclosed.
Anything factual you are asserting. If the video claims something happened, illustrating it with a generated depiction is a credibility problem, not a production shortcut. YouTube's disclosure rule covers AI that "Generates a realistic scene that didn't actually occur."
Screen recordings and demos. Record them. A generated interface is not the real product, and your audience knows the real one.
What YouTube says about disclosing AI
YouTube's help page says: "we require creators to disclose when they use AI to meaningfully alter or generate photorealistic content." You do it with the "AI use" setting in YouTube Studio. Its examples of content to disclose include "AI generated extra footage of a real place, like a video of a surfer in Maui for a promotional travel video", which is close to what realistic b-roll is.
What doesn't need disclosure, per the same page: "non-realistic content that's made with AI, or edits to realistic content that are minor", and "Production assistance, like using generative AI tools to create or improve a video outline, script, thumbnail, title, or infographic."
YouTube adds that "Disclosing AI content won't limit a video's audience or impact its eligibility to earn money." It may label content itself, including content that contains C2PA metadata, and creators who "consistently choose not to disclose this information may be subject to manual application of a label, or penalties from YouTube, including removal of content or suspension from the YouTube Partner Program." For ads and the legal side, see do you have to disclose AI-generated ads.
The cost of a video
8frame prices from its pricing code; a credit is about $0.01 at pack rate. Say a video needs eight pieces of b-roll, three cutaways and a thumbnail.
| Item | Model | Credits |
|---|---|---|
| 8 b-roll clips, 8 s, silent | Veo 3.1 Lite at 33 | 264 |
| 3 cutaways, 5 s, silent | Kling v3 Standard at 57 | 171 |
| 1 thumbnail | Nano Banana Pro at 21 | 21 |
| Per video | 456 credits, about $4.56 |
Kling v3 Standard needs a start image for each cutaway: a frame you already have costs nothing to use, and a generated one adds 6 credits on Seedream 5 Lite.
Over a four-week month, one video a week at 456 credits is about 1,820 credits: more than the $19 Starter plan's 1,000, inside the $49 Creator plan's 3,000. Two a week is about 3,650, past the Creator allowance. If 5-second b-roll clips are enough, the b-roll model sets most of the difference (same three cutaways and thumbnail in every row):
| 8 b-roll clips on | B-roll | Per video | Two a week, 4 weeks |
|---|---|---|---|
| Veo 3.1 Lite, 8 s, silent | 264 | 456 | about 3,650 |
| Wan 3.0, 5 s, 480p | 192 | 384 | about 3,070 |
| Hailuo H3 Max Turbo, 5 s, 480p | 136 | 328 | about 2,620 |
| Pruna p-video, 5 s, 720p | 112 | 304 | about 2,430 |
| Creatify Boreal, 5 s, 720p | 56 | 248 | about 1,980 |
| Pruna p-video Draft, 5 s, 720p | 32 | 224 | about 1,790 |
Every row except the first two fits two videos a week in the Creator plan's 3,000; for the first two, a top-up pack covers the difference. We have no quality measurements for these models, so generate one clip on the model you are considering and watch it in your edit before generating the rest.
Generate b-roll silent wherever the edit has music or narration over it: on Veo 3.1 Lite, silent is 33 credits against 54 with sound, about 40 percent less, and it is what you would mute anyway. Creatify Boreal has no silent setting; its sound is always on.
The workflow
Generate from a list. Before you generate anything, list what the script needs: that list is a shot list. Generating the clips in one sitting from the list, then cutting with the whole set in front of you, saves going back and forth between the edit and the canvas.
Reuse what you can. Generic b-roll (a city at night, hands on a keyboard, a coffee pour) can go back into later videos, so a library built over a few months lowers what each new video needs.
FAQ
Can AI generate a full YouTube video? Several tools will produce something. We don't recommend it for long-form, where people come for a presenter and an argument; use generation for b-roll, intros, cutaways and the thumbnail. If a video shows a real person saying something they didn't say, or a realistic scene that didn't happen, YouTube requires you to disclose it.
What is the best AI model for YouTube b-roll? We haven't measured which looks best, so here is the price side. On 8frame, Veo 3.1 Lite is 33 credits for 8 silent seconds at 720p. For 5-second clips, Pruna p-video Draft is 4 credits, Creatify Boreal 7, Pruna p-video 14, Hailuo H3 Max Turbo at 480p 17 and Wan 3.0 at 480p 24. Run the same prompt on two of them and compare the clips in your edit.
Do I need to disclose AI-generated b-roll? If it is realistic, yes: YouTube requires creators "to disclose when they use AI to meaningfully alter or generate photorealistic content," and its examples include "AI generated extra footage of a real place." Non-realistic content and minor edits don't need it, and neither does a thumbnail made with AI. Use the "AI use" setting in YouTube Studio; YouTube says disclosing won't limit a video's audience or its eligibility to earn money.
What aspect ratio should AI b-roll be for YouTube? 16:9, which YouTube's help page gives as the standard aspect ratio on a computer. On 8frame, the Veo, Kling and Wan nodes open at 9:16, so switch them; Kling v3 takes the shape of its start image.
What about Shorts? YouTube counts an upload as a Short when it is square or vertical and up to 3 minutes long. Different format, different answer: see best AI video generator for YouTube Shorts.
Sources
- YouTube Help, "Upload YouTube Shorts" ("Up to 3 minutes", "With a square or vertical aspect ratio"): support.google.com/youtube/answer/12779649, checked 2026-10-02
- YouTube Help, "Video resolution & aspect ratios" ("The standard aspect ratio for YouTube on a computer is 16:9"; avoid adding padding or black bars): support.google.com/youtube/answer/6375112, checked 2026-10-02
- YouTube Help, "Disclosing use of GenAI content" (what to disclose, what not, the "AI use" setting, automatic labels, penalties): support.google.com/youtube/answer/14328491, checked 2026-10-02
- 8frame prices: backend tool registry cost functions (calculateVeo31Cost, calculateKlingV3Cost, calculateWan3Cost, calculateHailuoH3TurboCost, calculatePrunaVideoCost, calculateCreatifyBorealCost; Nano Banana Pro and Seedream 5 Lite image prices), read 2026-10-02
- 8frame node defaults (Veo opens on Veo 3.1 Fast, 8 seconds, sound on; Veo, Kling and Wan nodes open at 9:16; Kling v3 needs a start image and takes its shape; Seedream picker labels): front-end tool registry and aspect-ratio config, read 2026-10-02
- 8frame plan allowances: Starter $19 for 1,000 credits, Creator $49 for 3,000, from the 8frame billing catalog
Generate the tissue, not the video.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.