Product Video Gen — Admin

How to use this tool

Quick walkthrough for marketing & ops users. No coding required — everything is in this browser tab.

What this tool does

Takes a Printerval product ID, a "vibe" (energetic, calm, futuristic…), a locale, and optional background music. Produces a finished short-form vertical video (TikTok / Reels / Shorts) with voiceover and a thumbnail. Drops the result in cloud storage and gives you a download button.

A typical run takes 3–6 minutes from Queue batch to ready. Queue several products at once and walk away — they process in the background.

1. Sign in

Two options on the sign-in screen:

• Client ID + secret — if the dev team gave you credentials, paste them here.
• Sign in with SSO — if shown, click and log in with your work account.

Stay signed in for ~12 hours (SSO) or longer (client creds). After that you'll be sent back to sign in again.

2. Workflows tab — the main thing you'll use

This is where you queue videos. The form starts with one empty Product #1 card. Fill it in, add more cards if needed, click Queue batch.

Adding more products: click + Add another product. Each one queues independently.
Removing a product: click Remove on the card.

Voice preferences (collapsed by default)

Click the Voice preferences strip to expand. Defaults work for most cases — only touch this if you want a specific voice.

• Voice — pick a specific named voice (e.g. en-US-Neural2-F). Locale must match the voice's language — don't pair an en-US-… voice with French.
• Gender — Female / Male / Neutral. Only used when you don't pick a specific Voice.
• Match vibe — auto-tweaks speaking rate / pitch based on the vibe (faster + brighter for "energetic", etc.). Leave unchecked for default delivery.
• Speaking rate / Pitch — manual overrides. Skip unless you have a reason.

Per-step provider overrides (collapsed by default)

Advanced. Lets you swap which AI handles each step. Leave alone unless someone tells you to change it. Defaults are tuned in the AI Config tab.

Or upload a CSV / Excel file

Below the form there's an "OR upload from spreadsheet" divider with a CSV upload area. For bulk batches (more than ~5 products at a time), this is the fast path — fill out a spreadsheet, upload, click queue.

• Click Download CSV template to get a starter file with the four columns: productId, vibe, locale, bgmUrl.
• Fill it in (one product per row). Save as .csv (preferred) or .xlsx.
• Pick the file with the Choose File input. The page parses it instantly and shows a preview table.
• Each row gets a status: ✓ ok (clean), ⚠ warn (some fields cleared, will still queue), or ⚠ skip (productId missing, won't queue).
• Click Queue N valid rows to submit. Same as the form — each row becomes one workflow.

Excel users: long product IDs sometimes get auto-mangled into scientific notation (2.43066e+09) when Excel "helps." If you see that in the preview, save as CSV instead — Excel preserves long IDs as plain text in CSV exports. The parser tries to recover scientific-notation values but it's not 100% reliable.

Audio — music, voiceover, clip audio

Three independent switches: on each Workflows product row, on the Meta ads tab, and as CSV columns music, voiceover, clipAudio (yes/no). Defaults: music on, voiceover on, clip audio off.

• Clip audio keeps the sound the video model generated. Only models marked 🔊 in the model picker produce any; the rest give silence.
• With clip audio on and voiceover off, each beat's script line is sent to the model so it speaks it itself.
• Nothing stops tracks clashing — music over the model's own music, or our voiceover over the model's speech. That is your call per job.
• On BytePlus Seedance, clip audio roughly doubles the price of each clip.

3. Job history & downloads

Every queued job shows up in the table below the form. Status flows: PENDING → RUNNING → COMPLETED (or FAILED).

• Filter by Status, productId, or change the limit. Click Refresh to reload.
• Click a row to expand it — you'll see each pipeline step's status, errors, and (for completed jobs) download buttons for the merged video, thumbnail, and individual clips.
• Downloads use temporary signed links that expire after ~15 minutes. If a link goes stale, just click again to get a fresh one.

4. Vibes tab — reusable mood prompts

A vibe is a short text description that the script generator weaves into its prompt — the difference between "calm meditation product showcase" and "high-energy hype reel".

• Existing vibes show in the table. Click to edit, trash icon to delete.
• Set or update a vibe — name it (e.g. holiday-cozy), give it 2–4 sentences describing the style + mood you want.
• Once saved, the new vibe is selectable from the autocomplete on the Workflows tab.

You don't have to create a vibe before using a name — unknown names get a generic default. But saving one gives you consistent output across runs.

5. AI Config tab — for advanced tuning

Most people will never need this tab. It's where the dev team and power users tune:

• AI step configs — which provider / model handles each pipeline step. Defaults work fine; only override if you have a reason. See "Picking a video provider" below if you're queueing kid products or licensed merch.
• Video gen settings — clip length and aspect ratio. Defaults: 8s and 9:16 (vertical). Cost estimate updates live as you change values.
• Base prompts — the master rules every AI call obeys. Each module has a built-in default; you can override and click Reset to revert. If you don't know what you're doing, leave these alone.

Picking a video provider

The tool doesn't pre-screen products — it just dispatches whatever you queue. Different providers have different content policies, so picking the right one for your SKU mix saves wasted runs. Approximate $/clip below is for a 32-second video (4 clips × 8s).

For licensed merch (band tees, world-tour shirts, official IP), assume Printerval already cleared the rights upstream — pick a provider that won't reject the request. Only Veo refuses pre-emptively.

Bull Board (link in the top nav)

A separate dashboard showing the raw queue state — pending jobs, retries, failures. Mostly useful for debugging when something goes wrong. Opens in a new tab and asks for a separate username/password (the dev team has it). You don't need this for normal use.

Common questions

A job failed — what do I do?

1. Click the failed row to expand it.
2. Look for the step marked FAILED and read its error message.
3. Common causes:
   • content filter refused — the AI blocked the request (usually a brand or copyrighted character it doesn't recognise as licensed). The product needs manual production.
   • Failed to fetch image … HTTP 403 — temporary storage hiccup. Re-queue the same product; usually works second time.
   • Anything mentioning a product field — the Printerval product ID might be wrong or the listing is missing data.
4. If the message doesn't make sense, copy the job ID and ping the dev team.

"United Kingdom" locale — what does the voice actually sound like?

Picks a British English (en-GB) voice. The Printerval region (/uk URL prefix) is unrelated — the locale only controls the voiceover language and the script.

Can I make videos for kids' products (onesies, baby tees, etc.)?

Yes — the tool now auto-detects when a product photo features a real child and switches the generated marketing scene to a people-free composition (flat-lay, studio backdrop). The product itself is the focal point, no model needed.

If a kid-product run still fails with content filter refused, the print on the garment itself probably depicts a child — those need manual production.

How long should a video take?

3–6 minutes from queue to download. Mostly waiting on the AI video generator. If a job is still RUNNING after 10 minutes, ping the dev team — something is stuck.

Can I pick the song from a library?

Not in the tool — marketing owns the licensed music externally. Either paste a direct .mp3 URL or upload the file. If you don't supply music, the video ships without a track.

How much does each video cost?

Open AI Config → Video gen settings — there's a live cost estimator that sums all the per-call costs (script + 4 images + 4 video clips + voiceover + thumbnail) for one workflow.

I want a voice that's not in the dropdown.

The dropdown lists curated Google voices that we know work well for Printerval-style content. Ask the dev team if you need something else — they can add it to the list.

Who to contact

• Tool is broken / down — dev team
• Want to add a new vibe / voice / locale — dev team (vibes you can do yourself in the Vibes tab)
• Output quality issues — start by tweaking the vibe prompt yourself; if that doesn't help, dev team can adjust base prompts