Grok Image-to-Video & Reference Images: A Practical Guide

TL;DR: Text-to-video is great when you're starting from words, but a lot of real work starts from images — a product photo, a character sheet, a logo, a set of brand elements you need to animate consistently. Grok Imagine's image-to-video, frame-to-video, and ingredients modes turn those images into motion, and the Grok Automation Chrome extension scales them by auto-matching each reference image to its prompt by filename. Name your files right once, and a whole batch attaches the correct image to the correct prompt with no manual uploading.

The three reference-driven modes

Grok Imagine has five modes; three of them take reference images as input. Choosing the right one decides what your clip is built from:

• Image-to-video — animates a single reference image according to your prompt. You supply the picture (a product shot, a character, a scene) and the prompt describes the motion or action you want applied to it.
• Frame-to-video — treats a supplied image as a specific starting frame and builds motion outward from it. Use this when you want the clip to begin from an exact image and then move.
• Ingredients-to-video — composes a clip from supplied reference elements ("ingredients") rather than a single source image, letting you combine inputs into one result. Like the others, it pairs references to prompts by filename for bulk runs.

All three share the same superpower in Grok Automation: reference images auto-match to prompts by filename. That one feature is what makes a 200-image animation job practical instead of a day of clicking "upload."

How filename matching works

Manually, the reference workflow is: open a prompt, upload its image, run it, repeat. That's fine for one clip and miserable for a hundred. Grok Automation removes the upload step by reading filenames. You put your reference images in a folder, you write a prompt list where each prompt corresponds to a reference filename, and the extension attaches the right image to the right prompt automatically across the entire batch.

The principle is simple: the name is the link. If your reference image and its prompt agree on a filename, they get paired. Get the naming consistent on both sides and the matching takes care of itself — there's nothing to drag, drop, or upload per item. That's also why a little planning of your folder and list up front pays off across the whole run.

A worked example: filenames to prompts

Say you're animating a set of six product shots for a launch. You drop six images into one reference folder and write a prompt list where each line corresponds to a reference filename. The pairing looks like this:

Reference filename	Prompt (the motion you want)	Mode
sneaker-red.png	Slow 360° turntable spin on a clean studio background, soft light	image-to-video
sneaker-blue.png	Slow 360° turntable spin on a clean studio background, soft light	image-to-video
watch-steel.png	Gentle push-in as light glints across the dial, shallow depth of field	image-to-video
bottle-amber.png	Condensation forms and a slow rise reveals the label	frame-to-video
logo-mark.png	Logo assembles from particles, then settles and holds	frame-to-video
brand-kit.png	Combine the supplied elements into a flowing brand montage	ingredients-to-video

Because the reference filenames and prompts are aligned, the extension attaches sneaker-red.png to its spin prompt, watch-steel.png to the push-in prompt, and so on — all in one run. Note the two sneaker prompts share identical motion text: that's deliberate, because keeping the prompt constant while varying the reference image is exactly how you produce a consistent set of clips for a product line.

Step-by-step: a reference-image batch

1. Gather and name your reference images. Put every source image in one folder and give each a clear, consistent filename that you'll mirror in your prompt list. Decide a naming convention now and stick to it.
2. Write the matching prompt list. Create one prompt per reference, keeping the filename correspondence exact. Reuse identical motion text wherever you want a uniform look across different images.
3. Install and open the extension. Add Grok Automation from the Chrome Web Store, open grok.com, sign in, and open the side panel.
4. Pick the reference mode. Choose image-to-video, frame-to-video, or ingredients-to-video depending on whether you're animating a single image, starting from an exact frame, or composing from elements.
5. Point the extension at your reference folder. Supply the folder of reference images so the filename matcher can pair each one with its prompt.
6. Paste your prompt list. Drop in the prompts you wrote to mirror the filenames. The queue auto-dedupes, and each prompt lines up with its matching reference.
7. Set output options and a download folder. Choose aspect ratio, resolution, and duration to match where the clips will be used, and set a filename template (timestamp plus prompt) so results arrive organized.
8. Run a small test first. Queue two or three pairs to confirm the right images attach to the right prompts. This is the moment to catch a naming mismatch cheaply.
9. Launch the full batch. Once the test pairs cleanly, run the whole list. Watch the live progress and activity log (with thumbnails), or walk away — auto-retry handles the occasional failure.

Structuring a folder and list that stay consistent

The whole system rests on disciplined naming. A few habits keep large reference batches reliable:

• One reference folder per project. Don't scatter source images; keep the batch's references together so the matcher has a single, clean place to look.
• Descriptive, collision-free filenames. Names like sneaker-red and watch-steel are self-documenting and unique. Avoid generic names like image1 that are easy to mismatch.
• Mirror the names in your prompt list. The prompt list and the folder are two halves of the same system — keep them in lockstep so every reference has a partner.
• Hold the prompt constant to vary the input. For product lines or character sets, reuse the same motion prompt across different reference images to get a visually consistent series.
• Vary the prompt to direct a single image. Conversely, point several prompts at variations of one reference when you want multiple takes on the same source.

Practical tips

• Test the pairing, not just the prompt. A small run confirms images attach correctly before you commit render time to the full set.
• Keep references in the right shape. Reference framing influences the output, so align your source images with the aspect ratio you'll generate at — see the aspect ratios & resolution guide.
• Lean on the activity log. Thumbnails of completed clips make it obvious at a glance whether the right reference drove each result.
• Let auto-retry cover hiccups. Set a retry count so a transient failure on one pair doesn't stall an unattended batch.
• Name outputs to match sources. A timestamp-plus-prompt filename template makes it easy to trace each finished clip back to its reference image.

Who this is for

Reference-driven batching is for anyone whose work starts from existing visuals: e-commerce teams animating product catalogs, brand and motion designers turning logos and kits into moving assets, character artists producing consistent action clips from a sheet, and marketers who need many variations that all stay on-brand. If you've been uploading one image at a time to animate a series, filename matching collapses that into a single run.

If you're new to the queue, begin with the beginner walkthrough, then read the bulk AI image & video workflow guide for prompt-list management, retries, and delays. Pair this with the aspect ratios & resolution guide to get the output shape right. For heavier use, see the Premium tier, or learn more on the Grok Automation home page.

Frequently asked questions

What is the difference between image-to-video, frame-to-video, and ingredients modes?

Image-to-video animates a single reference image according to your prompt. Frame-to-video uses a supplied image as a specific starting frame for the motion. Ingredients-to-video composes a clip from supplied reference elements. All three accept reference images and run in bulk.

How does Grok Automation match reference images to prompts?

It auto-matches references to prompts by filename. When you name a reference image to correspond to its prompt, the extension attaches the right image to the right prompt automatically across the whole batch, so you don't upload images one at a time.

How should I structure my reference-image folder and prompt list?

Keep all reference images in one folder with clear, consistent filenames, and write your prompt list so each prompt corresponds to a reference filename. Matching the naming on both sides lets the extension pair each prompt with its image and keeps a large batch consistent.

Do I need an API key or paid Grok plan to use the reference-image modes?

No. The extension drives the normal grok.com interface in your browser, so the reference-image modes work on any grok.com account including free Grok, with no API key and no separate signup.

What happens if a reference image has no matching prompt?

Only prompts paired with a matching reference filename are attached and run. Keeping your folder and prompt list aligned, and testing a small set first, ensures every reference is picked up before you launch the full batch.

Can I auto-download the resulting clips with organized names?

Yes. Every clip auto-downloads to a folder you choose using a filename template that can include a timestamp and the prompt, so reference-driven output arrives already named and easy to match back to its source image.

Is Grok Automation affiliated with xAI?

No. Grok Automation is an independent tool built to automate the grok.com interface. It is not affiliated with or endorsed by xAI. "Grok" is a trademark of xAI Corp.