Complete Guide

Everything you need to know about Auto Flow.

Create Image (Nano Banana)
  1. Select Create Image mode in the Control tab
  2. Paste one prompt per line in the Prompt List, or click Import file (.txt)
  3. Click Add to Queue
  4. Click Start — choose New Project or Current Project
  5. Monitor progress in the Task Progress table
  6. When done, switch to Gallery → Scan All to find your images
Input section showing Prompt List field, Queue Images with Shared Ref Upload and 1:1 Batch Upload buttons, bulk mode and repeat toggles, and Reference Library Input Section — Prompt List & Upload Options
Text-to-Video
  1. Select Text-to-Video mode
  2. Paste one prompt per line — each line is one scene with a clear camera direction
  3. Click Add to Queue
  4. Click Start and choose Create New Project or Run on This Page
  5. Watch Task Progress for Waiting → Generating → Done status

Best practices: keep one scene per line, put action + camera movement in the same line, use tags like [V1-S1] for stable sorting.

Frame-to-Video (Image-to-Video)
  1. Select Frame-to-Video mode
  2. Upload reference images using Shared Ref Upload or 1:1 Batch Upload
  3. Write video prompts in the Prompt List
  4. Optionally enable Map references by prompt to assign specific images to specific prompts
  5. Click Add to Queue → Start
Lower section of Control panel showing Reference Library, prompt mapping, Add to Queue and Manage buttons Reference Library, Prompt Mapping & Queue
Ingredients Mode
  1. Select Ingredients mode
  2. Upload product/ingredient images
  3. Write video prompts describing the product showcase
  4. Add to Queue → Start

Ingredients mode is designed for product showcase videos where you want to highlight specific visual elements.

Uploads & Reference Library

There are three ways to upload images in Auto Flow. Understanding the difference is important — only Add to Library saves images permanently.

Upload controls showing Shared Ref Upload, 1:1 Batch Upload, bulk mode toggle, and Reference Library Upload Controls & Reference Library
Shared Ref Upload

Upload reference images that are shared across all prompts in the current job. Every prompt in your list will use these images as visual references when generating.

These images are temporary — they are only used for the current job and are not saved to the Reference Library. If you clear the queue or start a new job, they're gone.

1:1 Batch Upload

Upload images that are paired 1-to-1 with prompts by line order. The first uploaded image goes with the first prompt, second image with second prompt, and so on.

Also temporary — these images are not saved to the Reference Library. They only apply to the current job.

Add to Library Persistent

Saves images to your Reference Library, which persists across sessions. Images saved here stay until you manually delete them. Use the library for images you reuse frequently — brand assets, character references, style guides, etc.

You can then assign library images to specific prompts using Map references by prompt.

Toggles

Bulk mode: 1 image per prompt (line order)

When ON, each uploaded image is assigned to exactly one prompt in order. Image 1 → Prompt 1, Image 2 → Prompt 2. When OFF, all images are used for every prompt.

Repeat first prompt for all uploaded images

When ON, the first prompt text is reused for every uploaded image. Useful when you have many reference images but want the same prompt applied to all of them.

Quick rule: Use Shared Ref or 1:1 Batch for one-off jobs. Use Add to Library for images you want to keep and reuse across multiple jobs.

Prompt Formatting Guide

How you format your prompts determines how Auto Flow processes them and how the gallery sorts your results.

Basic Format

One prompt per line. Each line = one generation task.

a golden retriever running on the beach, slow motion, cinematic
a cat sleeping in a sunbeam on a window sill
a hummingbird hovering near red flowers, macro shot

Tagged Format (Recommended)

Use tags like [V1-S1] to organize by video and scene. This enables proper sorting and Pick Matched functionality.

[V1-S1] a character walks into a dark laboratory, wide shot
[V1-S2] close-up of the character's face as they see the experiment
[V1-S3] the experiment glows bright blue, camera pulls back
[V2-S1] a chef prepares a meal in a busy restaurant kitchen
[V2-S2] close-up of the dish being plated with garnish

V = Video number, S = Scene number. This tells Auto Flow which prompts belong together. The gallery will group them and Pick Matched will pair images with their animated videos by matching these tags.

Multi-line Prompts

Separate prompts with a blank line if a single prompt spans multiple lines:

A detailed scene of a medieval castle at sunrise,
with fog rolling through the valley below.

A close-up of a knight's armor reflecting firelight,
cinematic depth of field.

Image-to-Video Format

When using Frame-to-Video with the ||| delimiter (for Auto Flow paste):

[V#-S#] image prompt ||| [V#-S#] video prompt

The left side generates the image, the right side animates it. Tags must match on both sides.

Full Example (2 videos, 5 scenes each)

Here's a complete real-world example showing how to structure a multi-video project:

[V1-S1] Neon city skyline in rain, reflective rooftops, dramatic clouds ||| [V1-S1] Cinematic tilt from skyline to street with rainfall streaking across lens.
[V1-S2] Courier under glowing signage, wet jacket, intense expression ||| [V1-S2] Slow dolly-in while steam drifts and signs pulse.
[V1-S3] Close-up data chip in gloved hand, water droplets, amber glow ||| [V1-S3] Macro rack-focus from droplets to glowing chip core.
[V1-S4] Narrow alley with hologram ads and mist, long perspective lines ||| [V1-S4] Handheld forward move through alley with subtle parallax.
[V1-S5] Rooftop dawn reveal, storm clearing over city skyline ||| [V1-S5] Crane-up shot as sunlight breaks through clouds.

[V2-S1] Desert outpost at sunset, dust haze, long warm shadows ||| [V2-S1] Slow panorama establishing outpost and surrounding dunes.
[V2-S2] Pilot inspecting drone wing by hangar, toolkit on ground ||| [V2-S2] Medium orbit around pilot with drifting dust particles.
[V2-S3] Cracked visor close-up reflecting lightning from distant storm ||| [V2-S3] Push-in close-up with subtle head turn and light flicker.
[V2-S4] Convoy vehicles driving into dunes, headlights cutting dust ||| [V2-S4] Side tracking shot matching convoy speed with motion blur.
[V2-S5] Night camp under stars, blue emergency lamps, quiet atmosphere ||| [V2-S5] Slow pullback to wide aerial while camp lights dim.

Download sample-prompts.txt

AI Prompting Guide

Use this prompt with ChatGPT, Claude, Gemini, or any AI to generate Auto Flow-compatible prompts. Save it as a custom instruction or paste it at the start of your conversation:

You are an Auto Flow prompt generator. You create prompts for bulk AI video and image generation. FORMAT RULES: - Each line is one prompt. One line = one generation. - Use [V#-S#] tags: V = video number, S = scene number. - For image-to-video: [V#-S#] image prompt ||| [V#-S#] video prompt - Image prompts describe the STILL FRAME (composition, lighting, subject). - Video prompts describe the CAMERA MOTION and ANIMATION (dolly, pan, tilt, zoom, tracking). - Keep each prompt on a SINGLE LINE (no line breaks within a prompt). - Be specific about: camera angle, lighting, mood, color palette, movement. - DO NOT add headers, comments, or blank lines between prompts. CAMERA VOCABULARY FOR VIDEO PROMPTS: Dolly-in/out, Pan left/right, Tilt up/down, Crane up/down, Tracking shot, Orbit/arc, Push-in, Pull-back, Handheld, Rack-focus, Slow zoom, Whip pan, Parallax, Aerial/drone. When I describe a concept, generate prompts in Auto Flow format. Output ONLY the prompts, no explanations.

Copy the prompt above and paste it into your AI of choice. Then just describe what kind of video you want and it will output ready-to-paste Auto Flow prompts.

Gallery Sorting Options

The sort dropdown controls how images and videos are ordered in the gallery:

Newest First Default

Most recently generated content appears at the top. Best for reviewing your latest batch.

1 → 99

Sort by prompt index ascending. Prompt 1 first, then 2, 3, etc. Follows the order you wrote your prompts.

99 → 1

Sort by prompt index descending. Last prompt first. Useful when your latest prompts are at the bottom.

A → Z

Alphabetical sort. If you used tags like [V1-S1], this groups all scenes of each video together in order — V1-S1, V1-S2, V1-S3, then V2-S1, etc.

Z → A

Reverse alphabetical. Last video/scene first.

Use A → Z sorting with tagged prompts like [V1-S1] for the cleanest gallery organization. All scenes from Video 1 will group together, then Video 2, etc.

Download Resolution (1K / 2K / 4K)

The resolution dropdown in the Gallery controls the quality of downloaded files.

Image Resolution

1K Default

Standard resolution (~1024px). Fast downloads, smaller file size. Good for previews, social media, or high-volume workflows where speed matters.

2K

High resolution (~2048px). Good balance of quality and file size. Suitable for most final outputs including prints and high-quality posts.

4K Pro

Maximum resolution (~4096px). Largest file size but highest quality. Best for large format prints, zoom-in detail, or archival quality. Requires Pro subscription.

Video Resolution

720p

Standard definition. Smaller files, faster downloads. Fine for social media where platforms compress anyway.

1080p (Upscaled) Default

Full HD. Upscaled from native resolution. Good quality for most uses.

4K Pro

Ultra HD. Maximum quality. Requires Pro subscription. Best for professional work and highest quality exports.

Higher resolution = larger file size = longer download time. For batch workflows with many images, start with 1K to preview, then re-download your picks at 2K or 4K.

Queue Management
  1. Build up multiple jobs by clicking Add to Queue multiple times
  2. Click Manage to view/edit queued jobs
  3. Set a custom Job Download Folder name
  4. Enable Auto-start next job in queue to chain them
  5. Click Start — Auto Flow runs each job sequentially
  6. Monitor progress with the live Task Progress table
Queue and progress tracking section Progress Tracking & Task Status Table

Format Guide

Structural tokens that tell Auto Flow how to split prompts, map references, and keep characters consistent. Descriptions can be in any language — the tokens stay the same.

Token Cheat Sheet
Scene tag — [V1-S1]

Optional, but useful for ordering and matching. V = video number, S = scene number.

Frame split — image prompt ||| video prompt

Left side creates the image. Right side animates it.

Character handle — @gary_before

ASCII handle, no spaces. Use underscores for display names with spaces.

Character fields — ||| voice: none ||| info: notes

Optional fields after the character description.

Character References (@handles)

Reuse the same character across scenes by giving it a handle once, then referencing that handle wherever the character appears.

  1. In Character Prompts, define each reusable handle once.
  2. In Character Sources, upload source images or pick existing project/reference images for each handle.
  3. In scene prompts, reuse the exact same @handle.
  4. For existing Flow characters, map the handle to the existing character instead of recreating it.

Syntax

Basic: @handle character description
Advanced: @handle character description ||| voice: Flow voice name ||| info: extra notes

@gary_before Overweight white man, plain white t-shirt, cheap athletic shorts, nervous expression. ||| voice: none @deandre Tall tan man, chiseled body, wavy dark hair, open shirt, smug relaxed face. ||| voice: Charon

Using characters in scene prompts

Use the same handle wherever identity matters. For long scripts, bind the important speaker/identity mentions first rather than every descriptive mention.

[Scene 2] Hotel hallway at night. @gary_before rushes forward and shoves @deandre against the wall. Dialogue: @gary_before, furious: "Who are you?" @deandre, calm: "Back up."

Do not use spaces inside handles. Use @gary_before, not @ gary before. A display name like “Gary Before” can still map to the handle @gary_before.

Auto Flow Packet Import

Use Paste Auto Flow Packet to bring characters, video prompts, and notes in as one block. Auto Flow previews the packet before applying it into the Character Prompts and Video Prompts sections.

# CHARACTER PROMPTS @test_luma Luma is a fictional young adult explorer with a teal jacket, silver backpack, short black hair, warm expression. ||| voice: none ||| info: curious, calm, adventurous @test_orren Orren is a fictional young adult inventor with an amber hoodie, round glasses, curly brown hair, thoughtful expression. ||| voice: none ||| info: clever, nervous, precise # VIDEO PROMPTS [V1-S1] @test_luma stands at the edge of a glowing forest path, teal jacket catching soft blue light, cinematic fantasy realism. ||| Slow dolly-in as @test_luma looks toward the trees. [V1-S2] @test_orren kneels beside a small device on a wooden table, warm workshop lighting. ||| Camera pushes in as @test_orren adjusts the device and glances toward @test_luma. # NOTES Style: cinematic fantasy realism, soft lighting, shallow depth of field. Continuity: Keep @test_luma in teal jacket. Keep @test_orren in amber hoodie and glasses.

Prompt for your LLM

You are writing an Auto Flow Packet. Use # CHARACTER PROMPTS, # VIDEO PROMPTS, and # NOTES. Define every @handle before using it. Use fictional non-famous characters. Use lowercase snake_case handles. Do not invent undefined @mentions. Use ||| to split the first-frame/image prompt from the video motion prompt. Use @handles only where identity matters.

After applying a packet, use Create / Verify Native Characters so Auto Flow maps each @handle to a real Flow character before generation.

Matching Modes
Flow Native @ Match Characters

For existing or created Flow characters. Uses Google's native character chip / entity binding path.

Auto Flow Classic Match Images

Auto Flow's original matcher for image references in Create Image, Frame-to-Video, and Ingredients. Maps uploaded refs by handle, filename, or prompt row.

Language-Agnostic Rules

Description text can be any language. Auto Flow only depends on the structural tokens:

  • Keep @handle in ASCII with no spaces.
  • Keep ||| as the split marker.
  • Keep optional field names as voice: and info:.
  • Everything after those tokens can be natural prompt text.

Start automating today

Join 60,000+ creators using Auto Flow.

Install Free Join Discord Telegram ★ Leave a Review