Series Lab Guide
A plain-language step-by-step guide: stages 1–7, a $0 quick start, the safety net, a glossary and FAQ.
What is Series Lab
Series Lab is a studio for creating a series of images (for example, an NFT collection) with one recognisable character or style. It works in two ways: on the website (sign in with a wallet, projects live on the server) and on your own computer (the local studio).
You set a base image and a set of traits (hair, outfit, background…). The studio makes a unique "recipe" for every NFT. Then an AI draws the images, or you stack them from prepared PNG layers. You pick the best ones and package everything for release.
Quick start in 10 minutes (mock · 1 cr)
Walk the whole path with the mock demo engine: generation is free, only building the ZIP at the end costs credits (1 cr, the price is shown next to the button). You will understand how it all works before using your balance.
- Open Studio in the browser and sign in with your wallet.
- 1 · Overview: name "Test", ID `test`, count 12 → "Save preset". Write the ID in Latin letters: letters, digits, dash.
- 2 · Base: upload any image with the role "Main base".
- 3 · Constructor: pick the pfp-core pack → "Apply pack" → "Save prompt+traits".
- 4 · Recipes: "Build texts".
- 5 · Generate: engine mock → "Pilot N" = 5 → "Estimate (free)" → "Generate…". You get 5 demo images.
- 6 · Review: mark ✓ / ↻ and try "Re-generate rejected" — v·2 versions appear.
- 7 · Export: "Run checklist" and "Build ZIP" (the only charge of this walk-through, usually 1 cr) — a package ready for a marketplace.
- This walkthrough does not mint. When the images are ready, build the deploy kit and, on this PC, follow the Deploy card: upload, hidden metadata, create the machine, mint, and reveal. Do not paste a key into the studio.
Mock draws simple coloured placeholders, not real portraits — it's a training ground. For real images pick a local or cloud engine.
Money: credits, top-ups, refunds
Paid actions use credits. You can see how many before you run.
- A new user gets welcome credits once, at the first wallet sign-in. There is no second free grant.
- Credits are spent by generation on cloud engines, by model vision (Vision) in Review and by the packaging stages (ZIP, deploy kit, marketplace kit: a small fixed fee, the price is shown next to the button). The "Estimate (free)" button on the "5 · Generate" tab shows the cost of the plan before you run it. Pricier engines (OpenAI, Kontext) cost more than the budget ones.
- To top up, open "Top up credits", pick a package and press "Pay with USDC". The bigger the package, the cheaper a credit.
- Payment is USDC on Solana, straight to us with no middleman. Press "Confirm in wallet": Phantom offers to send the exact amount. Credits arrive on their own, usually within a minute.
- Paid and nothing arrived? Wait a few minutes and open "Credit history". If there is no entry, write to support (the "Support and legal pages" section) with the short wallet address, the amount and the time of the transfer.
- If a frame could not be made, the credits come back: the history shows a "Refund" entry.
Step-by-step guide
Each stage is a tab at the top. Go left to right: 1 → 7, then Deploy and Market. You can go back at any time — nothing breaks.
1 · Overview — the series passport
Why: Sets the series name, how many NFTs there will be, and the random seed so the result can be reproduced.
What to do
- Enter the collection name and an ID in Latin letters (e.g. my-heroes).
- Set the number of NFTs — 12–24 for a first try.
- You can leave the seed as is. Remember: same seed = same characters.
- Pick an engine (changeable later) and click "Save preset".
What you get: The file presets/<ID>.json — your series settings.
Common mistakes
- An ID with spaces or Cyrillic — use Latin letters, digits and dashes.
- Changing the seed after review re-rolls the recipes and images stop matching their numbers. Change the seed only at the start.
2 · Base — the face of the series
Why: The base image is the model every NFT resembles. It decides whether the character is recognisable.
What to do
- Upload your image (role "Main base") or generate a candidate from a description.
- Look through the input/ gallery and make the best one the "Main base".
- Optionally click 🤖 Vision Assist to get a style description for the Constructor.
- That same main image is the bottom layer for "Compose layers" on Generate.
What you get: The file input/primary.png — the main base image and the bottom layer for PNG compose.
Common mistakes
- An image with text, logos or a busy background — the AI will copy those flaws.
- Several people in the frame — the series becomes inconsistent.
- A bright scarf or background in the base resists change: a «scarf colour» or «background» trait barely moves at any strength. Pick a base with a neutral background and without bright details you plan to vary. On a neutral base scarf colour obeys (green and red), but a strong background change does not.
3 · Constructor — the description formula
Why: Defines what stays the same in the images (keep, style) and what changes (traits).
What to do
- Apply a ready trait pack or load a preset.
- Check 3–6 categories that should vary.
- Review the options: value, weight (frequency), prompt (English words).
- Fill in keep and suffix, optionally Style lock. Click Validate and "Save prompt+traits".
- For prepared PNGs, make a ZIP with <key>/<value>.png folders and upload it on Generate in the "PNG layers" card (on a PC you can also put files in layers/<key>/<value>.png). The options table has no path field. The checkbox "This layer is not part of DNA" still paints the layer, but leaves it out of the trait fingerprint. A value of None paints nothing. Prompts stay.
What you get: A preset with the full trait dictionary and the prompt template.
Common mistakes
- Too many varying categories — images turn chaotic and the character gets lost.
- An empty prompt in an option — the AI won't know what to draw for that variant.
4 · Recipes — a plan for each NFT
Why: Turns the settings into an exact list: NFT #1 gets these traits and this description, #2 others, and so on.
What to do
- Click "Build texts".
- Check the table, the rarity report and twin warnings.
- Edit a single recipe if needed (click a row).
What you get: The archive out/<ID>.json with every recipe.
Common mistakes
- Changed the preset but didn't rebuild the archive — Generate uses the old recipes.
5 · Generate — drawing
Why: The AI draws images from the recipes. For AI images, credits are charged only when you click "Generate" with a paid engine. "Compose layers": a small per-frame fee (≈ 3 credits per 1000, at least 1 per run; previewing one frame and frames kept from an earlier run are free), the price is shown under the button.
What to do
- Pick an engine and reference strength.
- Choose "Pilot N" = 3–5.
- Click "Estimate (free)" and check how many credits will be charged.
- Click "Generate" and watch the progress. "Stop" aborts the stream; files already written stay. If nothing is left to generate, the stream does not start.
- If the traits are already drawn as PNG files, upload them as a ZIP in the "PNG layers" card (<trait>/<value>.png) and click "Compose layers". Or pick an already generated set in "Paint over": the layers are then painted over the model's frames and only traits that have a layer are painted. The base image is the main one from the Base tab. Prompts stay. A small per-frame fee: ≈ 3 credits per 1000 frames, at least 1 per run; the price is shown under the button, frames kept from an earlier run and a one-frame preview are free.
What you get: The folder out/<engine>-s<strength>/ with 1.png, 2.png… Earlier variants live in versions/. "Compose layers" writes out/layers and dna.json.
Common mistakes
- The whole series straight away without a pilot — you risk paying for hundreds of bad images.
- The chosen cloud engine is unavailable — try another or contact support.
- No base image or no layers (ZIP) — "Compose layers" stops. The card shows what is missing. The prompts stay either way.
- Scarf colour or background does not change — try the Kontext Pro engine (in live tests it held the scarf, background and expression best), lower the strength to 0.30–0.35 (for fal) or switch to a neutral base.
6 · Review — picking the best
Why: Not every image works out. Here you decide what goes into the collection.
What to do
- Pick a set, go through every image and mark ✓ or ↻.
- Redo the rejected ones — old variants remain as versions (v·N badge).
- In the versions window choose the best — "Make current".
- Delete erases the frame for good (the button asks for a second click) and it cannot be restored. Drop from set keeps the file in versions, but the frame leaves the set.
What you get: Marks (marks.json) and the final current versions of the images.
Common mistakes
- Images left unmarked — use the "Unmarked" filter.
- Delete erases a frame for good and it cannot be restored; Drop from set keeps the file in versions instead.
7 · Export — packaging for release
Why: Prepares files in a marketplace format. The sale is on the marketplace website. A Solana Core mint is a command on this PC after the deploy kit, not a button on this tab.
What to do
- Press «Run checklist» and fix what it flags.
- Pick a set, a format, and «Only keep-marked», then «Build ZIP».
- Optional: Arweave (Irys), «Provenance hash», «Showcase and rarity (HTML)», «LoRA training dataset».
- The Arweave button on this tab pays with the server key. A release upload is `series upload` with your key on this PC.
What you get: A ZIP with images and metadata for the marketplace. This archive does not start a mint.
Common mistakes
- Exporting without review — bad images end up in the collection. Turn on "Only keep-marked".
Deploy — a kit for your wallet
Why: Gives commands and a config. The machine is created with `mplx cm create`. You sign the Solana mint and reveal on your own PC. There is no on-chain deploy button in the studio, and you never paste a private key here. For EVM the reveal stays in the dashboard (Thirdweb).
What to do
- Pick a chain. The default is "Solana (Core Candy Machine, devnet first)". "Solana (Sugar, legacy)" is only if you deliberately stay on the old program. `series mint` builds a Core transaction only.
- Read the launch advisor and fill in the description and a public address. On Solana you can paste allowlist addresses, one per line. Press "Build deploy kit": the price of the stage is above the button.
- On your own PC make the release skeleton and upload the set: `series release-init <archive.json> -o release.json --creator <public address> --price <price>`, then `series upload <folder> --keypair <file>`. The key stays in the file.
- Create the machine with `mplx cm create` using the config from the kit, and check it with `series verify release.json`. Then mint (`series mint`) and reveal (`series reveal`). The full commands and flags are in `docs/RELEASE.md` in the repository.
What you get: A ZIP with DEPLOY.txt, an estimate from .env constants, and ADVISOR.txt. For Solana: cm-config.json, hidden.json, hidden.png, and allowlist.json when wallets are pasted. The hash in the kit is empty until the hidden-metadata step. Sugar lives in sugar/. After a mint, mints.json appears on this PC.
Common mistakes
- Do not paste a private key. The address field is a public address for the royalty share. The studio does not accept a key.
- Reveal reads mints.json, which `series mint` writes. A mint made by another program does not appear in that file on its own.
Market — a kit for the marketplace
Why: Prepares a checklist, listing copy, and images at a set size. You go to the marketplace yourself.
What to do
- Solana: Magic Eden and Tensor. EVM: OpenSea. Bitcoin: Ordinals marketplaces (not Magic Eden).
- Click "Build marketplace kit". The stage price sits above the button.
What you get: A ZIP with LISTING.txt, CHECKLIST.txt, and banner, avatar, and cover images.
Common mistakes
- Sizes in SIZES.json are the Lab slot defaults. Open the marketplace page before you print banners: the list can change.
Admin — resources, economics, errors
Why: Administrator-only panel: engines, spend, failures, trash.
What to do
- Admin → Resources: connection status. Admin → Errors: generation failures.
- Set the daily spend cap in server settings (see operator docs).
- Restore anything deleted by accident from the 🗑️ Trash.
What you get: Control over spending and files.
Common mistakes
- Changed server settings but didn't restart Studio — changes aren't picked up.
Hybrid and PNG layers
Hybrid is the default scenario: the AI paints the frames from the base image, and the traits that must be exact go on top as PNG layers.
- On the "1 · Overview" tab pick the "hybrid" scenario. The "Apply the hybrid recommendation" button sets Kontext and a reference strength of 0.35.
- On "3 · Constructor" mark a trait "Critical trait: painted from a PNG layer". The AI does not paint it, because in a prompt it holds weakly.
- Upload a ZIP of layers in the "PNG layers" card on "5 · Generate": one folder per trait, with <value>.png files inside.
- Generate the frames and press "Compose layers": the layers go on top of the model's frames. The source frames are not changed.
- The model may shift or rescale the face. The "Align the layer to the model frame" checkbox looks for the shift and scale from outlines. If the match is weak the layer stays put, and the message says why.
- Without a layer a critical trait will not appear on the frame. Before you generate, the studio warns you with the "Critical-trait layers are missing" dialog.
- A big series (hundreds of frames or more): «Compose layers» runs in the background, shows Composed N of M and has «Stop»; composing again continues instead of starting over. Before it starts the studio estimates the disk space.
Quality and review of frames
On the "6 · Review" tab you keep frames or send them to be redone. The "Series quality" card is free and runs locally.
- "Check quality" looks for broken files and exact copies and measures how close each frame is to the main base image.
- Similarity compares edges and composition, not colour and not the face. A low number means the frame deserves a closer look, not that it is bad.
- The card does not check that the requested traits are visible (background, scarf). Fix such a trait with a PNG layer.
- The button that marks the problem frames sets them to Redo and leaves your Keep marks alone. Then press "↻ Regenerate rejected".
- The "Fast Heuristics" auto-review is free, while "AI Vision" uses credits. A frame deleted in Review is erased for good.
- «Check quality» on a big set shows Checked N of M and has «Stop»; what was counted is kept.
Release: ZIP, Genesis, numbering
The "7 · Export" tab builds a ZIP for minting: images and metadata. It does not mint anything itself.
- Start with "Run checklist": it shows what is missing before release and gives warnings.
- Then "Build ZIP" and "⬇ Download ZIP". Files are named by edition number: 1.png, 1.json.
- If "Base image is part of the collection (edition #0)" is ticked on "1 · Overview", the base image becomes a separate lot #0 (Genesis), so the ZIP holds one edition more than the series has NFTs. Untick it to get exactly the number you set.
- "Name prefix" follows the series name until you type your own. "Image URI Prefix (Storage)" left empty gives plain file names; after the Arweave upload, put the address of the set there.
- For a real release, upload the set with your own key on your PC (`series upload`). Never paste a private key into the studio.
Licences and AI
A short note on rights. This is a guide, not legal advice.
- By uploading base images and layers you confirm you hold the rights to them. Under the Terms, Lab stores and processes them only to provide the service, including passing them to the AI provider you chose.
- The legal status of AI-made images depends on your country and on the terms of the engine itself (for example the FLUX model licences). Read the terms of your engine before a commercial release.
- The launch advisor ("8 · Deploy") has an "Image licence model" choice and an "Insert template" button. It is a draft for a lawyer, not a legal opinion.
- Say that images are AI-made wherever the marketplace or the law asks for it.
Safety net: how the studio protects your money and files
These work automatically — nothing to turn on (except the daily cap).
- Versions: re-generation never overwrites a paid image. Every variant is kept; switch the current one in Review (v·N badge).
- Trash: only whole projects have one. A frame deleted in Review or a dropped version is erased for good; other deleted and replaced files can be restored in Resources when you have access.
- Daily spend cap blocks a run that would exceed it before any paid generation (limit set by the administrator).
- Spend log: every paid call is recorded (time, engine, price, success or error).
- Auto-stop: if the engine fails several times in a row, the batch stops on its own.
- Estimate first: "Generate" is available only after the free "Estimate".
- Safety: payment details and service secrets never reach the browser; payment goes through your wallet.
- Brand filter: prompts with brand, franchise or celebrity names are blocked before generation.
How to work without spending credits
Credits are spent by cloud engines, by model vision (Vision) in Review and by stages 7, 8 and 9 (ZIP, deploy kit, marketplace kit). Everything else is free.
- Stages 1–4 and 6 never charge credits: recipes, selection with the Fast Heuristics and composing layers make no paid calls. Stages 7, 8 and 9 take a small fixed fee (usually 1–2 cr). The price is shown next to the button before you press it.
- The mock engine — demo images (0 cr), it spends nothing. Perfect for learning.
- ComfyUI / A1111 on PC: your GPU, a small Lab support fee from balance. See Local GPU help.
- Uploading your own base images is free; Vision Assist without AI runs a simple free analysis.
- Triage "Quick heuristic" is 0 cr; "AI Vision" spends credits.
- "Compose layers" on Generate costs pennies per frame (≈ 3 credits per 1000); only previewing one frame and continuing frames that are already done are free.
Glossary
Short and in plain words.
- Preset
- the series settings file: name, count, seed, traits, description template.
- Seed
- a word or number that fixes randomness. Same seed = same set of characters.
- Base image
- the model all NFTs in the series resemble. Main base = primary. The same file is the bottom layer for PNG compose.
- Trait
- a characteristic that varies: hair, eyes, outfit, background.
- Option
- a specific variant of a trait, e.g. "red hair".
- Weight
- how often an option appears. Weight 10 is ten times more often than weight 1.
- Trait pack
- a ready set of traits and options from libraries/traits/.
- Prompt
- the English text description the AI draws from.
- keep
- phrases present in every prompt (the fixed part).
- Suffix
- the prompt ending: quality, bans ("no text").
- Style lock
- fixed angle, light, mood and framing.
- Recipe
- the plan for one NFT: trait set + ready prompt.
- DNA
- a short fingerprint of the trait set. Same DNA = twins.
- Archive
- the file with all recipes of the series (out/<ID>.json).
- Engine
- the service or app that draws images (fal.ai, ComfyUI, mock…).
- Reference strength
- how closely an image sticks to the base: 0.1 free, 0.95 near copy.
- Pilot
- a trial generation of a few images before the whole series.
- Set
- the folder of images from one run, e.g. out/fal-s0.45/.
- Version
- one of the variants of an image for the same number. The current one is used by the gallery and export.
- Trash
- out/trash/: deleted projects, base images and replaced files. Restore them in Admin → Resources. Frames deleted in Review do not go to the trash.
- Daily spend cap
- the maximum spend per day (set by the Lab administrator).
- Review
- the tab where you keep frames or send them to be redone; automatic scoring gives a ✓ or ↻ suggestion.
- Upscale
- resizing an image to a larger size, for example twice as large.
- Metadata
- the JSON description of an NFT: name, image, traits. Marketplaces read exactly this.
- Arweave
- decentralised file storage; Irys — a service for uploading to it.
- Provenance hash
- a fingerprint of the collection proving the images did not change.
- LoRA
- a small fine-tune of an AI model on your images so it draws your character.
- Mint
- creating the NFT on a blockchain. For Solana Core that is `series mint` on this PC: the key stays in a file, not in a studio field. The sale stays on the marketplace.
- Launchpad
- a service for launching a collection sale. OpenSea Studio, LaunchMyNFT, and Magic Eden Creator Hub (Solana only).
- Gas
- the blockchain network fee for a transaction during mint.
- Hybrid
- a scenario where the AI paints the frame and the exact traits go on top as PNG layers.
- PNG layer
- a transparent image of one trait (a scar, glasses) that is laid over the frame.
- Critical trait
- a trait that must be exact: the AI does not paint it, only a layer does.
- Genesis
- edition #0: the base image of the series as a separate 1-of-1 lot.
- Devnet
- Solana's test network: minting there costs no real money.
- Similarity to the base
- a number from 0 to 1: how closely a frame matches the base image by edges and composition.
Frequently asked questions
The character looks different every time. What do I do?
Use a sharp base image; raise reference strength to 0.5–0.65; describe who the character is and the style in keep; turn on Style lock; reduce the number of varying traits.
How much does a series cost?
Click "Estimate" to see credits charged. Budget engines (fal, Pollinations) are usually 2 cr/image with a reference; Kontext Pro is 3 cr (it follows trait changes best); openai costs more. mock is 0 cr.
How do I avoid overspending?
Start with a 3–5 image pilot and watch the credit balance in the header. Free credits are only the one-time welcome grant; when they run out, top up. A wallet that has not paid or received a grant is also limited to a few generations per minute.
I re-generated an image and the old one was better. Is it gone?
No. In Review click the v·N badge on the image and pick the version you want — "Make current".
I deleted a base image by accident. How do I get it back?
Base images and replaced files stay in the Trash (Admin → Resources) for about 30 days: restoring needs administrator access, so contact support (see the "Support" section). A frame deleted in Review cannot be brought back.
It reports the daily spend cap and won't generate.
Today's spend cap is reached. Reduce the frame count or wait until tomorrow. The limit is set by the Lab administrator.
Generation stopped with a message about consecutive failures.
The engine returned errors several times (service unavailable or a temporary issue). Try another engine or contact support (see the "Support" section).
The "Generate" button is disabled.
Click "Estimate (free)" first. That's how the studio shows credits before charging.
The prompt is blocked because of brands.
The text contains a brand, franchise or celebrity name. Replace it with a generic description — this protects you from legal trouble on marketplaces.
Crooked hands and faces. What do I do?
Mark ↻ and redo; add "detailed hands, symmetrical face" to suffix; try another engine or strength and compare sets in Review.
Where are my files?
Presets, base images, archives and results live in the project folders on your PC (or on the Lab server). Image versions are in Review (v·N badge).
Why is a cloud engine I need missing?
The list of available engines is set by the Lab administrator. If something is missing — contact support (see the "Support" section).
Can I open the studio from a phone or another PC?
Yes — sign in with the same wallet on the Lab site. Local GPU (Comfy/A1111) only works on a PC where Studio runs next to the graphics card.
How do I train my own model (LoRA)?
Export → "📦 LoRA Training Dataset". Upload the ZIP of image + caption pairs to Kohya_ss or Replicate LoRA Trainer.
How do I go from the deploy kit to a mint and a reveal?
Build the deploy kit. On this PC, in this order: the release file (`series release-init`), upload (`series upload`), the hidden-metadata hash (`series hidden`), rules (`series guards`), the machine (`mplx cm create`), check (`series verify`), mint (`series mint --keypair`), and reveal (`series reveal --keypair`) with the collection update key. PNG layers are optional, via Compose layers. The allowlist is the Deploy field or release.json. The studio does not take a key. EVM reveal is a Thirdweb dashboard step.
How do I delete my data?
On the public site press “My data” in the top bar, read what will be deleted, type DELETE and confirm. Your files (base images, presets, images, layers, trash) and usage records disappear, then you are signed out. Only money records stay: balance, payments, debits. This cannot be undone.
How does inviting a friend work?
The top-up window (“+ Buy”) shows your “Invite a friend” link: copy it. When the invitee buys credits for the first time, you get a bonus (set on the server by `REFERRAL_BONUS_CREDITS`, 50 credits by default).
Where are the Terms and the Privacy Policy?
On the public Series Lab site: the `/terms/` and `/privacy/` pages (English: `/en/terms/` and `/en/privacy/`). In the studio the same paths lead to those pages. The sign-in page links to them. The service is in development and testing and provided “as is”; by signing in you accept the Terms.
I paid in USDC and got no credits. What now?
Wait a few minutes and open "Credit history". If there is no entry, write to support (the "Support and legal pages" section) with the short wallet address, the amount and the time of the transfer.
Can I sell the generated images?
That depends on the terms of the engine you chose and on the law in your country. Read the "Licences and AI" section and the engine's terms before a commercial release. We do not give legal advice.
Start and stop
Start: start-lab.bat (Windows) or npm start. Address: http://127.0.0.1:5344/studio/. Stop: stop-lab.bat or Ctrl+C in the server window. Switch the interface language with UK / EN at the top; the "Hints" toggle hides the purple hint boxes.