T2VA Shot Lists (No Reference Files)

T2VA is for “I only have words” — launch films, mood pieces, concepts before pack shots exist. You carry the full burden of specificity; there are no reference files to bail you out.

---

What H3 needs from you

Think storyboard + sound notes:

1. Shots with time — [Shot 1] 0–4s: or 0–4s: inline
2. Camera — push-in, truck left, static tripod, handheld micro-shake
3. Action you can see — “steam rolls across knuckles,” not “she feels cozy”
4. Sound in scene + music — separate blocks at the end

Official label for the main body: integrated_multimodal_description. You can use that header or plain [Shot 1] lines — both appear in working examples.

---

Camera verbs that actually register

| Say this | Instead of | |----------|------------| | Slow push-in toward hood badge | Cinematic reveal | | Truck left, subject static | Dynamic camera | | Locked tripod, no shake | Professional look | | Handheld micro-shake, documentary | Raw authentic vibe |

For one continuous take, write: “Single continuous shot, no cuts.” H3 may cut by default.

---

Dialogue in short clips

15 seconds fills fast. One or two short lines max:

(S1) says: "We're out of time." — controlled, not shouting.

Put dialogue in the same shot where the mouth is visible.

---

First / last frame modes (quick note)

• I2VA: “Treat @image1 as the opening frame; describe what changes forward.”
• FL2VA: “Path from @image1 to @image2 — continuous motion, no invented cuts.”
• L2VA: “Converge into the last frame from a plausible moment before it.”

FL2VA is not the place for montage unless you describe a continuous morph (rare).

---

Example skeleton (automotive hero)

integrated_multimodal_description:
[Shot 1] 0–4s: Wide — silver EV on rain-slick plaza, neon bokeh…
[Shot 2] 4–9s: Low tracking — silent launch, water spray…
[Shot 3] 9–12s: Rear — taillight bar, tunnel entry…

overall_soundscape: Rain, tire hiss, electrical relay clicks…
non_diegetic_music: Sub drone → pulse at 8s → silence last 2s.

---

When T2VA is the wrong tool

• You have accurate product photography → Ref2VA
• You need the exact face from a casting photo → Ref2VA
• You need logo-perfect label text → Ref2VA + explicit “lock label”

Browse T2VA commercial prompts — each breaks down structure before the full text.