TikTok is the platform where generated video is most likely to underperform, and the reason is cultural rather than technical: the format rewards specificity and presence, and generated content tends toward the generic. The accounts doing well with AI on TikTok mostly use it for the layer around a real person, not instead of one.
There is also a compliance dimension that other platforms handle more loosely, and it is worth reading directly rather than through any summary.
TL;DR
- A watermark is fatal here. It reads as low effort, so free tiers are for evaluation only
- Generate 9:16 natively, never crop. Ratio costs the same on every model
- Cheapest route: Wan 3.0 at 480p, 24 credits per 5 seconds
- Disclosure matters. TikTok has its own rules and labels for AI-generated and realistic synthetic content. Read the current policy on the platform
- What works: generation as b-roll, backgrounds and inserts around a real person on camera
Where generation helps on this platform
Cutaways and b-roll. The strongest use. A talking clip needs visual variation, and generated inserts at 24 credits per 5 seconds fill that cheaply.
Backgrounds behind text. Abstract motion under a caption sequence.
Concept and metaphor shots. Visualising something you are describing that you cannot film.
Volume variants for ads. For paid placements the calculation changes completely, because variant count is what wins there. See best free AI ad generators.
Where it does not
Replacing the person. TikTok's whole grammar is presence. A faceless generated feed competes on volume in a category where volume is free, which is a bad position.
Trend participation. Trends depend on specific sounds, specific formats, and being recognisably part of something. A generated approximation reads as missing the point.
Anything the audience can verify. Product, place, food, your actual life.
Disclosure, which is the part people skip
TikTok maintains its own rules for AI-generated content, including labelling requirements and stricter treatment of realistic synthetic media depicting real people or events. These have changed more than once and vary by region.
Read the current policy in the app or on TikTok's own site before you build a format on it. We are not going to summarise numbers or thresholds here that may have moved; the honest advice is to check the source. What has been consistently true: synthetic media showing real people or presenting invented events as real is treated seriously, and undisclosed realistic synthetic content is the category most likely to be actioned.
What it costs
Canon 8frame prices, a credit is $0.01 at pack rate:
| Route | Clip | Credits |
|---|---|---|
| Wan 3.0 at 480p | 5s | 24 |
| Wan 3.0 Prime, any resolution | 5s | 34 |
| Veo 3.1 Lite, silent | 8s | 33 |
| Veo 3.1 Lite, with audio | 8s | 54 |
| Kling v3 Standard, silent | 5s | 57 |
| Gemini Omni Flash, with audio | 8s | 135 |
| Assembly | Merge Videos | 10 |
Twelve cutaways for a month of posting: 288 credits, about $2.88.
The format that works
Film yourself for 20 seconds saying one specific thing. Cut in three or four generated inserts where you would otherwise hold on your own face. Caption it properly. Post consistently.
That is unglamorous and it is what the data supports: the real person carries the account, and generation removes the friction of finding visuals to cut to.
FAQ
Can I post fully AI-generated videos on TikTok? Subject to TikTok's current AI content policies, including labelling. Check the platform's own rules, which change.
Does AI content get suppressed? Platforms have been tightening guidance on mass-produced and low-effort content generally. Interchangeable is the problem more than synthetic is.
What resolution should I generate? 720p is fine for most placements, and generate vertical natively rather than cropping.
What is the cheapest way to add b-roll? Wan 3.0 at 480p, 24 credits for 5 seconds.
Be on camera, generate the cutaways. The 8frame canvas is free and unlimited, and generation is paid from $19/month.