TikTok 'Reference to Video' vs Sume frame_images for product moments

TikTok's newsroom says Reference to Video in Symphony Creative Studio lets advertisers prompt the exact images and products they want at specific moments. On Sume, the closest tools are frame_images, which pin an exact first or last frame, and input_references, which give the model visual guidance but not exact frames.

The TikTok line is from TikTok World '26, dated May 13, 2026 and read on 2026-10-07. The same page says Dreamina Seedance 2.0 was integrated into Symphony. We did not test Symphony.

Side by side

The TikTok column uses only the page's wording; the Sume column is from the video guide.

What Sume can pin, and what it cannot

The Sume docs describe pinning only the first and last frames. A product that must appear in the middle of the clip cannot be pinned there with frame_images. There are two workable ways:

• Split the story into short jobs. Each job starts on the product moment you want, and you join them with a timeline render.
• Use input_references and a prompt that names the moment, then check the result and rerun if it misses. Nothing guarantees the product appears at a given second.
• For a last-frame product shot, use last_frame so the video ends on the pack.

Check before you spend more

Pull stills at the moments that matter with video inspect frames: { at: [1.5, 4.0] } and compare them to your product image. Because no model takes a seed, a good take cannot be reproduced; keep the file. The poll response shows usage.cost, so you know what each rerun cost.

Which to pick

If your ads are made and bought inside TikTok, Symphony keeps the work in one place. If you need other models, an API, or the same product frames in ads for other platforms, Sume gives you the frame controls above. For the input comparison in general see Symphony prompts versus Sume inputs, and for the field precedence see frame_images and input_references together.