RelayMe 如何協助市場營銷團隊將一張產品相片製作成廣告

Disclosure: I work on RelayMe, a self-funded product based in Hong Kong. This article describes how the product works today. Pricing and model availability can change — check the live site before you budget a campaign.

Marketing teams rarely lack ideas. What they lack is time between “we need a new angle for this SKU” and “the creative is ready for Meta, TikTok, and the PDP.”

A familiar loop: one approved product photo sits in Drive. Someone opens three AI tools — studio still, vertical crop, short clip — then pastes outputs into Slack and the ad account. Credits, logins, and “which model did we use last week?” pile up. The photo is fine. The process is not.

RelayMe is built around a narrower bet: keep the product photo as the source of truth, wire the steps once on a visual canvas, and price each model in credits before you run. Image, video, and text nodes share one balance. The output of one step can feed the next.

The problem is fragmentation, not “AI quality”

Most teams already tried standalone generators. Image quality can be good enough for tests. The friction shows up elsewhere:

• Tool sprawl — stills, upscales, and video in separate tabs
• Opaque cost — you learn what a clip cost after it finishes
• Inconsistent handoff — the 1:1 hero and the 9:16 story ad diverge after the first edit
• Hard to redo — when the brief changes, you rebuild the chain from memory

A workflow canvas does not magically fix brand risk or model drift. It does make the pipeline visible: input → candidates → upscale or outpaint → video → download. Swap the photo; keep the same settings.

One photo, four practical steps

1. Start with an Image input

Upload a product photo you have rights to use. Keep label text, silhouette, and color as the reference for every later check. Treat AI as a production assistant, not as the brand owner of record.

2. Generate studio or lifestyle candidates

Connect an image model — for example Nano Banana Pro or GPT Image variants in the selector — and describe the scene. Generate a small set of candidates (1–4 per image task), then compare each against the original: label spelling, proportions, packaging color.

Do not batch ads until at least one still passes that check. Upscale and video will amplify whatever is already wrong.

3. Upscale or outpaint for placement

Approved stills often need a higher tier (2K / 4K) or a different aspect ratio. Outpaint to 9:16 for Stories and Reels; keep square or 4:5 for feed. Upscale and Outpaint are separate modes on supported sale-channel models — the selector shows which options apply.

4. Animate with a video node that accepts a first frame

Use the approved image as the first frame (or reference, depending on the model). Pick duration, resolution, and audio, then review the credit estimate before you run.

Video models currently listed include Veo (Veo 3.1 / Veo 3.1 Fast), Grok Imagine, and MiniMax (MiniMax-H3). Image options include GPT Image family models and others on the pricing page. Availability is whatever the live selector shows. RelayMe hosts a canvas that calls third-party models; it does not train its own.

Review motion the same way you reviewed the still: product shape, labels, and whether the motion fits the channel.

Wire those four steps once. Next SKU: replace the input photo and rerun.

Why credit pricing matters for marketing ops

Ad creative is a volume game with a cost ceiling. RelayMe prices generation in credits: you see the estimate for the model, settings, and output tier before a run. Subscription credits refill monthly and do not roll over; separately purchased top-ups do not expire. Trial credits, when an offer is active, follow the validity shown at sign-up.

Public plans as of this writing (USD, monthly):

Images are often tens to a few hundred credits each; video is priced per second (often from around 100 credits/s depending on model and settings). A short clip usually costs more than several stills — so stills first, then animation.

No vanity metrics: your cost is whatever the estimate and pricing table say for that run. RelayMe is self-funded and based in Hong Kong (support Mon–Fri, 09:00–17:00 UTC+8).

Good fit — and limits

Good fit: e-commerce teams producing many variants from a small photo library; performance marketers testing hooks without a full studio day; agencies that want a reusable chain; anyone who wants one bill across image + video steps.

Not a substitute for: guaranteed pixel-perfect packaging (always QC labels); loading local LoRAs or custom checkpoints (not supported); unlimited free generation; automatic commercial rights beyond applicable terms — check RelayMe’s terms, the model provider’s terms, and your own input rights before publishing ads.

A simple team checklist

1. Lock one reference photo per SKU.
2. Generate 2–4 still candidates; keep only matches.
3. Upscale / outpaint for each placement.
4. Animate only the winners; compare models if needed (Veo vs Grok Imagine vs MiniMax on the same first frame).
5. Export, QC labels again, then upload to the ad platform.
6. Save the workflow so the next product does not reinvent the graph.

Try it without rewriting your stack

Register without a credit card when an eligible trial is active; the intro credit amount and expiry are shown at sign-up. Start with image nodes so you learn the QC loop before spending heavier video seconds. Pricing, model list, and credits are on the public pages at relayme.ai.

Marketing creative will keep changing. The useful constant is a pipeline you can inspect, price, and rerun — starting from the one product photo you already trust.

Questions about plans or models? Email support@relayme.ai.