Guide
How to make a TikTok product video with AI (step by step)
A good TikTok product video used to mean a shoot: the product, lighting, a phone on a tripod, and a few hours of takes. With AI video you can compress that into a single photo and a prompt — and end up with a vertical 9:16 clip that has motion, a setting, and sound. This is the exact workflow we use in Renoise to turn one product photo into a TikTok-ready video, start to finish. It runs on Seedance 2.0 (ByteDance) by default, with Kling 3.0 Omni (Kuaishou) for talking, UGC-style clips where someone speaks on camera.
Before you start
Two things make this go smoothly. First, a clean product photo — straight-on or three-quarter, on a plain or uncluttered background, well lit. The model anchors the video to whatever you give it, so a sharp reference beats a busy one. Second, a rough idea of the shot: where the product sits, how the camera moves, and the vibe of the setting. You don't need a script; you need a description.
Everything below happens on the Renoise Canvas, where the models, your uploaded materials, and the generation controls live in one workspace.
The workflow
1. Gather your product material
Upload your product photo to the material library on the Canvas. This is the asset pool the model will draw from. Once it's in, you can reference it directly inside a prompt by typing @ and picking it — that tells the model "this exact object goes in the shot," instead of hoping a text description lands the right look.
You can add more than one material if it helps: a second product angle, a logo, a background plate. Seedance 2.0 accepts multimodal references — up to 9 images, 3 video clips, and 3 audio tracks in a single generation — so you have room to anchor color, shape, and even a soundtrack reference if you want one.
2. Write the prompt
Describe three things: the shot, the motion, and the setting. A workable pattern:
@my-producton a marble kitchen counter, soft morning light, slow push-in as steam rises behind it, shallow depth of field.
Naming the product with @ keeps it locked to your reference. The rest — camera move, lighting, environment — is what turns a static photo into a video. Keep it concrete: "slow push-in," "handheld pan," "product rotates on a turntable" all read better to the model than "make it cinematic." Seedance 2.0 is web-search enhanced, so brand or style cues it can ground itself with also help.
3. Pick a model and mode
For most product videos, stay on Seedance 2.0 and use image-to-video (also called first-frame): your product photo becomes the opening frame, and the model animates outward from it. That's the mode that keeps the product looking like your product.
If you want a person talking about the product — a UGC-style testimonial or a creator-to-camera hook — switch to Kling 3.0 Omni. It has native lip sync, so a spoken voiceover matches the on-screen mouth movements, plus multi-subject consistency for keeping a presenter and product coherent across the clip. If that presenter is a real person or a spokesperson, use authorized real faces — Renoise gates real-likeness footage behind a one-time likeness review, so you're only animating faces you're cleared to use.
| Goal | Model | Mode |
|---|---|---|
| Product hero shot, motion + setting | Seedance 2.0 | Image-to-video (first-frame) |
| Spoken UGC / talking presenter | Kling 3.0 Omni | Lip sync with a voiceover |
4. Set 9:16 and let the model add audio
TikTok is vertical, so set the aspect ratio to 9:16 before you generate. Seedance 2.0 supports six aspect ratios (21:9, 16:9, 4:3, 1:1, 3:4 and 9:16), and 9:16 fills the TikTok frame edge to edge with no letterboxing or awkward cropping later.
Set the clip length anywhere from 4 to 15 seconds — a TikTok product hook usually lives in the 6–10 second range. One more thing that saves an editing pass: Seedance 2.0 generates audio, so you get ambient sound or a soundtrack baked into the clip instead of a silent video you have to score afterward.
5. Iterate, then extend
Your first generation is a draft. Read it back against the prompt: did the camera move the way you asked, is the product on-model, is the pacing right? Tighten the prompt — sharpen the motion verb, adjust the lighting, swap the reference — and regenerate. Drafting at lower resolution keeps iteration cheap while you dial it in.
When a clip is right but you need it longer, use continuation to extend from the last frame instead of stitching a new clip onto the end. That keeps the motion and the setting continuous, which matters when a 6-second hook needs to run to 12 for a fuller beat.
6. Export watermark-free
When you're happy, export. Watermark-free export is a paid-plan feature in Renoise, so a clean, ready-to-post vertical file is what you hand to TikTok — no overlay to crop around. Bump the resolution to 1080p (or 4K on Seedance 2.0) for the final render once the draft is locked.
What it costs
Renoise is credit-based across every model — you spend the same credits whether you're on Seedance 2.0 or Kling 3.0 Omni. At the lowest per-unit rate that comes out to AI videos from $0.34 per video and AI images from $0.03 per image, so a short product clip is a few cents of compute, and iterating a couple of times is still cheap. Plans run from Lite ($5 / 300©) up to Advanced ($200 / 14,000©); there's no free tier, but a single Lite top-up covers a lot of drafts.