Bulk image & video generation on magnific.com, straight from a side panel.
Magnific Automation turns a list of prompts into a queue and runs it on magnific.com for you — submitting each prompt, waiting for the result, and downloading it. You choose the model and Magnific-side settings (resolution, aspect ratio, and so on) on Magnific itself; the extension generates with whatever you have selected there and handles the repetitive work.
Four steps, in this order. Most problems on a first run come from step 1.
The side panel shows a short How it works guide when you first open it. You can reopen that guide from the Setting tab; this page is the fuller reference.
| Magnific page | Modes that run there |
|---|---|
| Image Generator | Text → Image, Image → Image |
| Video Generator | Text → Video, Image → Video, Frame Pair → Video |
If you see Wrong page, there are two equally good fixes: open the page that matches your mode (the message has a button for it), or pick a mode that matches the page you are already on. Switching modes back and forth without changing the page will keep producing the same message.
Pick a mode in the Generation mode board at the top of the Control tab.
Prompt → video. Each line of the prompt box becomes one video. Nothing to upload.
Image + prompt → video. Each uploaded image becomes the start frame of one job, paired with its prompt: image 1 with prompt 1, image 2 with prompt 2, and so on.
Not every Magnific video model accepts a start frame, and the same model can change what it accepts when you change its settings. The extension reads the live page before running; if the current model/settings have no start-frame slot, Run stops with Start frame not supported. Choose a model/settings combination that shows a start-frame upload on Magnific.
Start frame + end frame + prompt → video. Upload your images, then choose a Frame pair logic for this batch — there is no default, on purpose:
The Frame pairs preview shows every pair before you run. More on Adjacent vs Grouped →
This mode needs a model/settings combination that shows both a start and an end frame slot. If only a start frame is available, Run stops with Start and end frames not supported — use Image → Video instead, or switch to a model with both slots.
Prompt → image. Each line becomes one generation. Outputs / Prompt (1, 2 or 4) sets how many images each prompt produces — the extension sets Magnific's own count control before each submit, so the side panel's choice wins. On Unlimited, Magnific may cap this at 1.
Source images + prompt → image. For each job the extension clears the previous
references, uploads the new ones into Magnific's References area, reads the
@img1, @img2 … label Magnific gives each one, and includes those
labels in the prompt it submits. Your original prompt is what you see in the queue and report.
References per prompt (1–6) is the maximum number of images one job may use. With Auto-attach images off, images are grouped in order: with 2 per prompt, images 1+2 go to prompt 1, images 3+4 to prompt 2, and a final odd image still forms a valid job on its own.
Turn Auto-attach images on to match by filename instead: an
image is attached to any prompt that mentions its name (without the file extension, ignoring
case and punctuation). blue-wolf.png matches "draw the blue wolf". Matching is
whole-name, so anna.png will not attach to "annabelle". Very short names are
ignored, because a file called a.png would match nearly every prompt. Every
prompt must match at least one image, or Run stops with
Match an image to every prompt; images no prompt mentions are
reported as unused.
Each line is one prompt; blank lines are ignored, so you can space things out. A prompt cannot span several lines. The count above the prompt box shows how many prompts were actually parsed — check it before you spend anything.
In image-driven modes, Reuse applies one prompt to every uploaded image or frame pair. It is available only when exactly one prompt is entered, so extra prompts can never be silently ignored.
Enter a number next to the prompt count and click Jump to select that prompt without scrolling. Clear empties the prompt box; for eight seconds afterwards, Undo restores exactly what was there. Typing or importing new text cancels the Undo so it can never overwrite new work.
In image-driven modes, Prompt matching shows every job's prompt next to the exact image, start frame, end frame or references that will be submitted with it. Use it to catch ordering mistakes before any credits are used.
Images are paired with prompts by position: prompt 1 uses the first image, prompt 2 the second, and so on (Image → Image with Auto-attach on is the one exception — it matches by filename). So the order of your images matters.
These live in the Control tab's Configuration block.
Choose Yes or No before your first run — the extension deliberately does not guess. With Yes, each finished image or video is saved the moment Magnific finishes it; with No, results stay in your Magnific feed and you download them yourself.
001_, 002_ …
(or 001_v1_ for multiple outputs) so files sort in prompt order even when
Magnific finishes them out of order. Off by default.
Creates a real subfolder inside your Downloads folder — by default
Downloads/magnific-batch/. Non-English folder names work. To change where your
browser saves downloads in general, use Browser settings under
Download Settings in the Setting tab.
On by default. Every batch ends with a CSV with one row per prompt: the prompt, its status, a timestamp, and output URLs, plus diagnostic columns that tell you what happened:
| Column | What it tells you |
|---|---|
| Failure Stage | submit — never reached Magnific, nothing was generated.
rejected — Magnific's moderation refused it.
wait — it may have generated; check your Magnific feed.
download — it generated and is in your feed; only the file is missing.
Blank means the prompt succeeded. |
| Expected / Actual Outputs | How many outputs were requested vs. produced. |
| Submit Retries / Resolve Retries | Submit retries clicked Generate again. Resolve retries only kept waiting for the same generation — they never charge twice. |
| Downloads OK / Downloads Failed | Reported separately, so "nothing attempted" is distinguishable from "everything landed". |
| Card IDs, Output URLs, Requested Filenames | For matching files to Magnific cards. The filename is the one requested; your browser may add " (1)" if a file already exists. |
The maximum number of generation jobs in flight at once, while keeping prompt order. For video, if prompt 1 asks for 4 outputs and this is 5, the extension submits prompt 1 four times and prompt 2 once; each completed job frees a slot. For images, Magnific's own output-count control makes multiple outputs from one submission. Higher concurrency is faster but puts more load on your Magnific account — lower it if your model is rate-limited.
Before each submit, the extension waits a random number of seconds in your range. The minimum is 5 seconds so batches never hammer Magnific. Increase it if you see throttling.
Images use Magnific's own count control. Video modes produce more than one output by submitting the same job repeatedly. Unlimited may cap this at 1.
Retries, Wait Limits, Screen lock and your preferences are in the Setting tab:
Magnific's moderation occasionally refuses a harmless prompt. Retries on Block (on by default) tries it again; Maximum Retries (1–5, default 1) counts retries after the first attempt. With retries off, each prompt gets one attempt.
How long to wait for each generation before giving up. Both default to Auto.
Runs one prompt and one output at a time, start to finish (it sets Concurrent Prompts and Outputs / Prompt to 1). Use it if results are being marked Failed while the image is actually sitting in your Magnific feed: with only one generation in flight, no other generation can be mistaken for it. It does not help with slow queues, moderation, or a background tab. The extension suggests Safe mode at the end of a batch where this happened.
The Setting tab sets the extension's language (also in the header) and the Default Mode shown when the panel opens. The extension's language only changes the extension — it never changes Magnific.
Browsers give inactive tabs fewer resources — this is not an extension issue — so automation in a background tab may not work correctly. To work on something else meanwhile, keep the Magnific tab visible in its own window or on a second monitor. The side panel shows batch progress and reminds you to keep Magnific visible on the first accepted batch of each day. Turn on Screen lock while running in Setting if you want to cover the Magnific page and prevent accidental clicks during a batch.
Run automation is always clickable. If a batch is already running, the new one is queued and starts when the current one finishes. Each batch is a group in the Prompt queue showing every prompt's status (Queued, Submitted, Retry, Done, Failed, Cancelled) and a countdown to the next prompt. Remove takes a waiting batch out before it starts; Clear finished tidies up completed ones. You can keep editing the form while a batch runs — each Run takes a snapshot, so edits only affect the next batch.
Stop means "no more new submits". Generations already submitted keep running on Magnific and still download — Magnific has no way to cancel a generation once it starts, and the credit is already spent. Prompts not yet submitted are cancelled.
Afterwards a resume bar offers Resume (continue where you stopped) and Start fresh. Resume is offered only while your prompts and images are still the ones from that batch. The main Run button always starts a fresh batch.
The batch stops and your place is saved, with one of these messages: The Magnific tab reloaded, The Magnific tab left the generator, or The Magnific tab was closed — each followed by "Your place was saved - click Resume to continue." Moving around inside Magnific's generator does not stop a batch.
If three prompts in a row are submitted but nothing appears at all, the extension stops with Stopped: nothing generated for 3 prompts in a row. This usually means Magnific is rate-limiting, your Magnific session expired, or the page changed. The remaining prompts are not attempted, so you don't wait out every timeout for nothing. Moderation refusals and download failures don't count toward this.
After an image batch with timed-out prompts, the extension makes a final pass over your Magnific feed and may recover results that finished after their waits ended. Video batches do not have this final recovery pass; check the feed if a video times out.
Magnific controls Unlimited — the extension never turns it on or off. It only reads the live page: when Unlimited is on, the panel shows a warning, and the first Run shows Unlimited mode is ON. If you still want to continue, acknowledge the notice and click Run again. Magnific's Fair Use Policy
When Run can't start, the extension shows one of these. The bold text is exactly what you'll see.
| Message | What to do |
|---|---|
| Wrong page | Your mode needs the other Magnific page. Open the page your mode needs, or pick a mode for the page you're on. See the table. |
| No active tab | Open Magnific in a browser tab, then run again. |
| Magnific page not ready | Reload the generator page, wait for it to finish loading, then click Run again. |
| Automation unavailable | The extension couldn't load its compatibility settings. Try Check for fixes (see Getting help); if it keeps happening, check your network or contact support. |
| Sign in again | Your extension session couldn't be verified. Sign out, sign back in, then try again. |
| Message | What to do |
|---|---|
| Add a prompt | Enter at least one prompt. |
| Upload a source image | This mode starts from images — upload at least one. |
| Upload frame images | Frame Pair → Video needs at least two images. |
| Choose pair logic | Select Adjacent or Grouped for Frame Pair → Video. |
| Match an image to every prompt | Auto-attach is on and some prompts mention no uploaded filename. Rename an image so its name appears in the prompt, or turn off Auto-attach images. |
| A prompt/image count mismatch | Add or remove prompts or images until the counts match. With exactly one prompt, turn on Reuse to use it for every image. |
| Message | What to do |
|---|---|
| Choose auto-download | In Configuration, choose Yes or No for Auto-download. It's saved for future runs. |
| Choose file naming | Auto-download is on, so choose Native or Prompt under Filename source in the Control tab's Configuration block. |
| Message | What to do |
|---|---|
| Start frame not supported | The current video model/settings have no start-frame slot. Choose a model/settings combination with one. |
| Start and end frames not supported | The current model/settings lack an end-frame slot. Pick one with both, or use Image → Video. |
| Cannot verify frame inputs | The extension couldn't read the model's frame slots. Reload the video generator page, then try again. |
| Unlimited mode is ON | See Unlimited mode. |
| What you see | What it means / what to do |
|---|---|
| Marked Failed, but the image is in your Magnific feed | The result was generated but couldn't be matched to its prompt in time — most likely with several prompts running at once. Turn on Safe mode for that batch. |
| Generated, but no file in Downloads | Check Auto-download
is Yes. In the Debug Logs tab, a saved file shows as a ↓ Saved line
((URL fallback) means the extension downloaded it directly when Magnific's
download button didn't appear). The CSV's Downloads Failed column counts misses. |
| Files land in Downloads, not your folder | Check Save to folder has a name, and that your browser isn't set to ask where to save each file. |
| Everything stopped after a few prompts | See Automatic stop after repeated empty results. |
| Automation suddenly stopped working after a Magnific update | Click Check for fixes — compatibility fixes are delivered without a store update. If it still fails, report it (below). |
Magnific Automation is preconfigured browser automation built for Magnific. It does not ask ChatGPT, Claude, Codex, Manus, Perplexity or another AI agent to look at the page and decide every click, so you don't spend separate AI-agent credits on repetitive browser work. Magnific generations themselves still use your Magnific credits or plan.
Permissions, and why each is needed:
storage | Save your settings. |
tabs | Tell whether the active tab is a Magnific generator page. |
sidePanel | Show the extension as a side panel. |
scripting | Connect to Magnific tabs that were already open when the extension was installed or updated, without reloading them. |
alarms | Keep the extension awake during a batch and refresh compatibility settings hourly. |
identity | Sign in with Google. |
downloads | Save results into a named subfolder and name them. |
Compatibility settings come from magnific-automation.pages.dev. If your network
blocks it, the extension can keep using a structurally compatible cached configuration. Without
one, it shows Automation unavailable and won't run. A live response that
excludes your extension version also blocks automation, even if a cached copy exists.
Settings are stored in your browser (and synced by your browser if you use browser sync). Your sign-in tokens stay on your device. Your prompts, images and generated results are exchanged only between your browser and magnific.com; the extension's own servers are used for sign-in and compatibility settings. Full details are in the privacy policy.