Product Ad Video

Turn a product into a short ad clip that keeps the product exact. Start from a good still, move the camera around it, and add sound. One call makes one shot; cut shots together in an editor.

Inputs to collect

• Product image: a public https URL or a TokenLab result URL. If the user has no clean shot, make one first with tokenlab-product-photography. Several angles help reference-to-video models.
• Brief: the one thing the viewer should notice (texture, size, the lid opening, steam, a pour).
• Format: aspect ratio per channel (9:16 Reels/Shorts/TikTok, 1:1 or 4:5 feed, 16:9 web) and length per shot.
• Sound: none, ambient sound and effects, or a short spoken line.
• Hard rules: product shape, color, logo, and on-pack text must not change. Add no claims, prices, or badges the user did not supply.

Models

<!-- tokenlab:models:start --> <!-- Generated from the live catalog by scripts/sync_models.py. Do not edit; change preferredModels in catalog.json. -->

Operation	Model	Pick	Use
image-to-video	veo3.1-fast	default	Animate one product still
image-to-video	veo3.1	alternative	Higher tier (not live-tested)
image-to-video	kling-v3.0	alternative	Alternative; sound on or off (not live-tested)
image-to-video	grok-imagine-video	alternative	Low-cost draft
reference-to-video	seedance-2.0	default	Keep the product from several photos
reference-to-video	veo3.1-fast	alternative	References, alternative (not live-tested)
reference-to-video	grok-imagine-video	alternative	Low-cost references
text-to-video	veo3.1-fast	default	Brief only, no photo
text-to-video	seedance-2.0	alternative	Brief only, alternative (not live-tested)
text-to-video	grok-imagine-video	alternative	Low-cost draft

This list refreshes daily from the live catalog. Preferred models come first; otherwise models are ranked by their 24-hour success rate. Before a call, confirm the model with GET /v1/models/{model} and read its price from GET /v1/models/{model}/pricing.

<!-- tokenlab:models:end -->

How to choose

Exact fields and values come from each model's contract.

• One product still (image-to-video): Veo 3.1 Fast is the default, with sound, 4, 6, or 8 s at 720p or 1080p. Veo 3.1 is the higher tier. Kling 3.0 is an alternative with sound on or off. Grok Imagine video makes a low-cost draft.
• Several product photos (reference-to-video): Seedance 2.0 keeps the product from several photos. Veo 3.1 Fast is an alternative; it takes 8 s only for this operation. Grok Imagine video makes low-cost reference drafts (1-10 s).
• Brief only, no photo (text-to-video): Veo 3.1 Fast, then Seedance 2.0 as an alternative. Grok Imagine video makes a low-cost draft.
• Sound: Veo always adds audio. For silent shots, pick a model whose contract lists output_audio.

All video calls use POST /v1/videos/generations with an explicit operation. Send output_audio only when the operation lists it.

Workflow

Run each call with tokenlab-run (contract → estimate → call → poll → result). Commands below use its helper, ../tokenlab-run/scripts/tokenlab_run.py, a path relative to this skill's folder. If you installed this skill without tokenlab-run, send the same requests with curl. --wait stops after 100 s; a longer task keeps running, so continue with the resume command it prints.

1. Get a clean still. Use the user's photo if it is sharp and well lit. Otherwise run tokenlab-product-photography first and keep its result URL as the anchor.

2. Read the contract: python ../tokenlab-run/scripts/tokenlab_run.py contract MODEL --operation image-to-video (or reference-to-video, text-to-video). Use only listed fields and values.

3. Estimate the cost: python ../tokenlab-run/scripts/tokenlab_run.py price MODEL, then per-second price × duration for each shot. Some models charge more with audio on or at a higher resolution. Use the Official price as the upper bound. Tell the user the total for all shots before you start.

4. Build one body per shot:

   {"model": "MODEL_ID", "operation": "image-to-video", "image_url": "https://...", "prompt": "Slow push-in on the white ceramic mug on a walnut table. Soft morning light moves across the glaze. Coffee steam rises. Quiet room tone, a soft clink. The mug shape, color, and logo stay exactly the same.", "duration": 6, "resolution": "720p", "aspect_ratio": "9:16"}
   

   Take MODEL_ID from the Models list above for the shot's operation. Set duration and resolution to values its contract lists.

5. Call and wait: python ../tokenlab-run/scripts/tokenlab_run.py run /v1/videos/generations --json @shot1.json --wait. Save each task ID. Poll every 5-10 s.

6. If a wait times out, resume with python ../tokenlab-run/scripts/tokenlab_run.py poll TASK_ID. Never create a duplicate task.

7. Check each shot against the Quality bar. Re-run only the weak shot, after you tell the user the cost.

8. Return each video URL with its model, task ID, aspect ratio, and estimate.

Technique

• Restage, then animate. Fix the product still first. A video model copies the flaws of its first frame.
• Slow, simple camera moves sell products. Use one move per shot: slow push-in, slow orbit, low-angle tilt up, top-down reveal. Fast moves blur labels.
• Light is the motion. "A light sweep moves across the glass", "sunlight shifts across the table" add life without changing the product.
• Protect the product in words. End the prompt with what must stay: "the bottle shape, label, and cap stay exactly as in the image."
• One action per shot. Pour, open, rotate, or steam. Make a second shot for the next action.
• No generated text. Do not ask the model to write slogans or prices on screen. Add text, logo cards, and prices in an editor.
• Sound. Veo models always add audio. Describe it: "soft clink of the cup, quiet morning room tone". Put a spoken line in quotes. For silent shots, pick a model that lists output_audio and send false.
• Multi-angle consistency. For a hero product shown from several angles, use reference-to-video with 2-4 clean product photos instead of separate image-to-video calls.
• One aspect ratio per call. Make 9:16 and 16:9 as separate calls. For image-to-video, crop the still to the target shape first, because some models follow the input image shape.

Quality bar

• The product matches the input: shape, proportions, color, logo, and label.
• No warped text, extra parts, or floating products. Shadows follow the light.
• Each shot has one clear action and a steady camera.
• Aspect ratio and length match each channel. No invented claims, prices, or badges.
• The user saw the total estimate before the jobs, and the answer lists every shot's model and task ID.

Related skills

• tokenlab-run: the call, poll, and result mechanics.
• tokenlab-product-photography: make the clean product still first.
• tokenlab-animate-image: generic image-to-video with no product rules.
• tokenlab-cinematic-video: camera language for brand mood shots.
• tokenlab-talking-avatar: a presenter who talks about the product.
• tokenlab-voiceover: a separate voice track for the edit.