Prompt Guide

Everything you can put in a prompt — with copy-ready examples.

TryAIToday Flow Automator runs a list of prompts on Google Flow in one batch — for images or videos. Beyond plain text, you can pull in reference images, reuse an image by name with @tag, name a new image with #name so later prompts can reuse it, and — in the Video tab — drive each clip with @start / @end frame tags.

Two tabs, same idea. Switch between Image and Video at the top of the side panel. Each keeps its own prompts, queue and history, so you can set up one without disturbing the other. Sections 1–5 below apply to both; section 6 covers what's specific to video.

Basics Reference images @tag — reuse an image #name — name an image Chaining (consistency) Video prompts Tips

1. The basics

Type or upload one prompt per block. Separate each prompt with a blank line — every block becomes one generation (or several, if you set images-per-prompt above 1).

a cozy cabin in a snowy forest, warm light in the windows

a neon city street at night, rain reflections, cinematic

a studio portrait of an astronaut, soft key light

Load a .txt file or paste directly, pick your model / aspect ratio / quality once in Settings, then press Start. Results preview live in the queue and auto-download into a per-run folder.

2. Reference images

In the Reference Images section you can add Subject (who / what), Scene (where), and Style (look / mood) images — from your Flow Library or a quick Upload. They apply to every prompt in the batch.

Up to 5 per slot, and 10 in total across Subject + Scene + Style combined.
Use these when the whole batch should share the same look or characters.

3. `@tag` — reuse an existing image by name

If an image in your Flow project has a name, drop it into a prompt with @ and it's pulled in as a reference for just that prompt:

@lina standing in a misty forest, golden hour, wide shot

@name for single-word names; @[two words] for names with spaces.
The name is the image's name in Flow. An uploaded file like lina.png works as @lina (the extension is optional — @[lina.png] also works).
If several images share a name, the newest one — or the one you just made in this run — is used.
An @tag the extension can't find is simply left in the prompt as plain text.

4. `#name` — name the image a prompt creates

Add #name to a prompt to give the image it generates a name. The #tag isn't part of the picture — it just labels the result (and renames it in Flow) so you can reference it later with @name.

a young woman, red hair, freckles, studio portrait #lina

#name or #[two words], same as tags.
One name per prompt. When a prompt makes several images, the name is given to the first one.

5. Chaining — the same character across a batch

Combine the two: name a character once, then reuse it in the prompts that follow — all in a single run, no need to refresh anything.

a young woman, red hair, freckles, studio portrait #lina

@lina sitting in a sunlit cafe, candid, 35mm

@lina hiking a mountain trail at sunrise, wide cinematic shot

@lina in a spacesuit on a red-rock planet, dramatic light

The first prompt creates and names lina; every prompt after it reuses that exact image as a reference — so it's recognizably the same person throughout.

You can mix both in one line — e.g. #hero @city walking out of the station at dusk names the new image hero while referencing your city image.

6. Video prompts

Switch to the Video tab to batch-generate clips the same way you batch images — one prompt per block, separated by a blank line. Pick your model, aspect ratio, videos-per-prompt and duration once in Settings, choose a video mode, then press Start.

Video modes

Text — prompt only; the model creates the clip from your words.
Start frame — the clip begins from an image you choose and animates forward.
Start + End frame — you give the first and last frame; Flow interpolates the motion between them.
Reference — up to 4 reference images guide the subject / look, and the clip is generated to match.

Frame tags — `@start`, `@end`, and references

Just like @tag for images, you can name the exact frames a single video should use, right inside its prompt. Which tags are available depends on the mode:

Start frame: @start:name sets that prompt's opening frame.
Start + End frame: @start:name and @end:name set the first and last frames.
Reference: @name (plain, up to 4) pulls images in as references — same syntax as the Image tab.

@start:sunrise the camera pushes slowly through the misty valley, cinematic

@start:door_closed @end:door_open the old wooden door swings open into daylight

Single-word names use @start:name; names with spaces use @start:[my hero].
Names resolve to a Flow image exactly like @tag / #name in the Image tab — so an image you named earlier (or uploaded) can become a start or end frame.
Tags are removed from the text sent to the model, and shown as chips on each queue row so you can see which frames a clip will use.

Different frames for each prompt

Prefer picking visually? Open Different for each prompt to assign a start / end / reference image to every prompt from a grid — no tags needed. Anything you set there fills in for prompts that don't already carry a tag.

Precedence: an @tag in the prompt wins, then a per-prompt assignment, then the shared frame set in Settings. So you can set one default and override just the prompts that need something different.

Models & durations

Omni Flash — fastest; supports 4 / 6 / 8 / 10s and Text, Start-frame, and Reference modes.
Veo 3.1 Lite / Fast / Quality — higher quality at 8s; Lite and Fast support all modes, Quality is best for Text and frame modes.
The panel auto-corrects impossible combinations (for example, resetting the duration to 8s for Veo models) and tells you why.

Finished clips preview in the queue and auto-download into their own Videos run folder, separate from your images.

7. Tips

Keep each prompt on its own block, separated by a blank line.
Set Subject / Scene / Style references for a look that applies to the whole batch; use @tag when only one prompt needs a specific image.
Name early: put #name on the prompt that establishes a character or product, then @name in the rest.
Names are case-insensitive and can use letters, numbers, - and _ (e.g. @product_01).
Click any finished thumbnail in the queue to preview it full-size right away.
For video, name an image with #name in the Image tab first, then use it as @start:name or @end:name over in the Video tab.

Questions or ideas? Join our Discord or email codewithsiddique@gmail.com.