QuickStart Guide
Last updated
Was this helpful?
Last updated
Was this helpful?
This QuickStart walks you through generating AI images and minting NFTs using ChainGPT’s API and SDK — with real examples, valid parameters, and everything you need to go live fast.
Before you begin, make sure you have the following in place:
ChainGPT API Key: Sign up on the ChainGPT platform and obtain an API key from the API dashboard. Every request must include this key in the Authorization header (as a Bearer token). Also ensure your account has sufficient credits (see Pricing & Credits below) – requests will fail if you run out of credits.
Node.js Environment: Install (for using the SDK). You’ll also need a package manager like npm or Yarn.
API Access Tool: For making raw API calls, you can use cURL (as shown in examples) or any HTTP client (Postman, etc.). All API endpoints are relative to the base URL https://api.chaingpt.org.
Secure Key Storage: Treat your API key like a password – store it securely (for example, in environment variables or a secrets manager), not hard-coded in your code or pushed to repositories.
ChainGPT’s AI NFT Generator API allows you to generate AI images from text and optionally mint them as NFTs. Below we demonstrate the workflow using direct REST calls with cURL.
To generate an image from a text prompt, use the POST /nft/generate-image endpoint. This returns the image data directly in the response, suitable for quick previews or standalone image generation.
Request: Include your API key in the header and provide the prompt and generation parameters in JSON. For example:
In this request:
prompt – The text description of the image to generate.
model – The AI model to use (e.g. "velogen", "nebula_forge_xl", "VisionaryForge", or "Dale3" for DALL·E 3).
steps – (Optional) Number of refinement steps. Higher values can yield more detail but take longer. Each model has a range: velogen supports 1–4 steps (default 2), nebula_forge_xl and VisionaryForge support up to 50 (default 25), while Dale3 (DALL·E 3) doesn’t use a steps parameter.
height and width – The image dimensions in pixels. Ensure they are within the supported range for the chosen model (see Model Options below). Common base sizes are 512 or 768 for quick generation, or 1024 if supported.
enhance – (Optional) Upscaling option for higher resolution. Use "1x" for 1× enhancement or "2x" for 2× enhancement. If omitted or set to "original", the image is generated at base resolution (no upscale). Using enhancement produces a larger image (up to the model’s max upscale) at additional credit cost.
Response: On success, you’ll get a JSON response with the image data. For example:
The data field is an array of bytes representing the generated image (in this case, a JPEG). You can convert this to an image file. For instance, in Node.js you could do:
This will write the bytes to output.jpg. (If the API returns a base64 string instead, decode it accordingly.)
Model Options & Parameters: The API supports multiple models, each with different capabilities:
velogen – Fast generation for smaller images. Supports base resolutions up to ~768px and 1–4 steps (quick iterations).
nebula_forge_xl – High-quality model for larger images (up to 1024px) and more steps (up to 50 for fine detail).
VisionaryForge – Similar to Nebula Forge XL (up to 1024px, 50 steps), with its own style/profile.
Dale3 – Integration of OpenAI’s DALL·E 3 model. Fixed 1024×1024 output, no custom steps or enhancements allowed. Use this for DALL·E’s unique output, but note it has a separate pricing structure (see below).
Each model has recommended resolution ranges. For example, velogen is optimal at 512–768px (can upscale to ~1920px with enhance: "2x"), whereas Nebula/VisionaryForge excel at 768–1024px (up to ~1536px with upscale). If you request a resolution beyond a model’s base capability without setting enhance, the API may automatically upscale it if possible. It’s best to explicitly use the enhance parameter if you want high resolution output.
If you intend to mint the image as an NFT, it’s recommended to use the queued generation endpoint. This allows the backend to handle longer-running jobs and prepare NFT metadata. Use POST /nft/generate-nft-queue to start the generation without waiting for the image immediately.
Request: In addition to the prompt and image settings, you must provide:
walletAddress – The blockchain wallet address that will ultimately receive/own the NFT.
chainId – The identifier of the blockchain network where the NFT will be minted. (Use the Get Chains endpoint or see below for how to find supported chain IDs.)
Example cURL to queue a generation:
This request queues an image generation with the given parameters (prompt, model, etc.) on BSC Mainnet (chainId: 56). Higher values for steps (like 25 or even 50 for Nebula) can yield very detailed images, and here we used enhance: "2x" for an upscaled 1024→1536px output.
Response: The server immediately returns a job identifier:
The collectionId is a unique UUID for this generation task. The status will initially be "queued" (or "processing" shortly after) – no image is returned yet. You will use this ID to query the generation status and later to mint the NFT.
After queueing an NFT generation, you can poll its status using GET /nft/progress/{collectionId}. This returns the current status and (if completed) the result of the generation.
For example:
Response: If the job is still in progress, you’ll see something like:
Here, status is "processing" and progress is ~45%. (Some jobs may not return a percentage, just a status.) When the status becomes "completed", the response will include the image data:
In the above completed example, data contains the bytes of the generated image (starting with 137 80 78 71 which is the PNG file signature). You can now save or utilize this image (just as in step 1).
Often, instead of manually retrieving the raw image here, the next step is to call the minting API to get the NFT metadata. The image is stored on the backend once generation completes. For simplicity, you can proceed directly to Mint the NFT after you see a "completed" status.
Tip: You don’t have to query progress in a tight loop. Polling every few seconds is sufficient. Alternatively, you could implement a webhook/callback if the API provided one, but polling is straightforward for most use cases.
If you have a short or less-detailed prompt and want to improve it before generating an image, you can use POST /nft/enhancePrompt. This endpoint uses AI to refine your prompt (making it more descriptive or creative), which can lead to better image outcomes.
For example:
Response:
The API returns an enhancedPrompt field containing a more detailed version of your input. You can then take this suggestion and use it as the prompt in your next generate-image or generate-nft-queue call. Each prompt enhancement call costs a small number of credits (0.5 credits per call).
(Using prompt enhancement is optional but recommended if you’re not getting the desired detail in the outputs. It can help maximize the quality of the image, especially for expensive models or high-step generations.)
The AI NFT Generator supports minting on multiple blockchain networks. To see which chainId values you can use, call GET /nft/get-chains:
By default, this returns only mainnets. Add ?testNet=true to include test networks. A sample response:
Each entry has a chainId and name. Use these IDs in the generation and minting calls. For example, 56 for Binance Smart Chain, 1 for Ethereum, 137 for Polygon, etc. If you’re testing on a testnet (like BSC Testnet or Goerli), be sure to both use the testnet chainId and a wallet address from that test network.
Once your image is generated (and you have a collectionId from the queued generation), the final step is to mint the NFT. This involves calling POST /nft/mint-nft with the collectionId and some NFT metadata. This call will prepare the NFT’s metadata (and potentially trigger the on-chain mint, depending on the setup).
Request: Provide:
collectionId – The ID from the completed generation job (ensure the status was “completed” before minting).
name – A name/title for the NFT (e.g. "Neon Skyline City").
description – A description for the NFT, detailing the artwork or any info you want in the token metadata.
symbol – A short symbol or ticker for the NFT/collection (e.g. "NEON").
Example cURL:
Response: If successful, you will receive metadata about the newly created NFT. For example:
This includes the name, description you provided, and importantly an image URL (often an IPFS link to the image that was uploaded). Additional metadata like attributes could appear if relevant. The transaction field may contain a transaction hash if the system initiated an on-chain mint for you, or null if not applicable.
At this point, the NFT’s data is ready. If the service did not automatically mint the NFT on-chain, you have two options to finalize the minting:
Use the ChainGPT NFT contract to mint the token to the provided wallet address (you'll need on-chain interaction, see Contract ABI below).
Or, if transaction was returned, wait for that transaction to confirm on-chain (or fetch it via a blockchain explorer).
In most cases, you will need to perform the actual blockchain transaction yourself. The mint-nft response gives you the metadata (and the image is now pinned e.g. on IPFS). Now you or your backend can call the blockchain smart contract’s mint function to create the NFT.
If you plan to interact with the NFT smart contract directly (to manually call the mint function, or to integrate with web3 tooling), you can fetch the contract’s ABI via GET /nft/abi. This returns the ABI (Application Binary Interface) JSON of the ChainGPT NFT Mint Factory contract.
Example:
Response:
This ABI JSON can be used with a web3 library (Ethers.js, Web3.py, etc.) to invoke the contract. Typically, the NFT contract has a mint(address to, string tokenURI) function you’ll call, passing the walletAddress and a token URI (which the backend has likely already created for you, pointing to the metadata JSON). The ABI also defines events like NFTMinted that you could listen for.
Using the ABI from the API ensures you’re interacting with the correct contract and function signature. After calling mint-nft (Step 6), you would use this ABI along with the collectionId/metadata info to execute an on-chain transaction to actually mint the NFT to the user’s wallet.
ChainGPT provides an official JavaScript/TypeScript SDK (@chaingpt/nft on npm) to simplify integration. This SDK wraps all the above API calls into convenient functions. In this section, we’ll accomplish the same tasks using Node.js.
First, install the SDK into your Node.js project:
Import and initialize the SDK in your code with your API key:
Here we store the API key in an environment variable for safety (you can replace process.env.CHAINGPT_API_KEY with your key string for testing, but avoid hardcoding secrets in production). The Nft class instance nft will be used to call all NFT API methods.
To generate an image, use the nft.generateImage() method. This mirrors the POST /nft/generate-image endpoint.
A few notes on the above code:
The parameters prompt, model, steps, height, width, enhance are the same as described in the REST API section and are passed in as an object. The values here will produce a 512×512 image using the velogen model with 2 refinement steps and a 1× enhancement (effectively upscaled from 256px base to 512px output, since velogen’s base resolution is smaller).
imageResult will be an object containing the image data. Typically, the SDK returns a structure similar to the API’s JSON. In this case, imageResult.data.data is the byte array of the image (the double .data is because the SDK may wrap the response). We convert it to a Buffer and write to a file.
Error handling: we wrap the call in a try/catch to catch any exceptions (network issues, invalid parameters, etc.).
This method is synchronous from the caller’s perspective (it awaits the image generation to complete and returns the image). Use this for quick generation or when you need the image immediately.
Model options and parameters (as discussed earlier) apply here as well. For example, you could switch model to "nebula_forge_xl" and increase steps to 25 for a higher-quality image (just ensure to adjust height/width within that model’s range, e.g. 1024×1024, and consider using enhance: "2x" for maximum quality). The SDK will call the appropriate API and retrieve the result.
You can retrieve the list of supported blockchains via the SDK using nft.getChains(). By default, getChains() returns mainnets; pass true to include testnets.
This will fetch the same data as the /nft/get-chains API. Each call returns an object containing a chains array (accessible via .data in the SDK response). For example, chainsMainnet.data.chains might include entries like { chainId: 56, name: 'BSC Mainnet', ... }. Use these IDs for the chainId parameter in generation and minting.
To generate an image with the intent to mint it as an NFT, you have two options in the SDK:
Immediate generation with nft.generateNft(): This will wait for the image generation to complete and return the image data plus a collectionId. Use this if your application can afford to wait (e.g. 30-60 seconds) for the generation to finish in one go.
Queued generation with nft.generateNftWithQueue(): This will return immediately with a collectionId (similar to the REST generate-nft-queue), and you can then poll progress with nft.getNftProgress() if needed. Use this if you want to handle the waiting or progress feedback in your app manually (for example, show a progress bar).
For simplicity, here’s how to use generateNft() for a synchronous flow:
This call will take some time to resolve (it internally queues the generation and waits for completion). Once it returns, you get an object result containing the image data and a collectionId. We log the collectionId for reference. At this point, the image is generated and stored, and you can proceed to mint it.
If instead you prefer to not block execution, you could use generateNftWithQueue():
This returns immediately with the collectionId. You can then call nft.getNftProgress({ collectionId }) in intervals to check status and progress, similar to the REST polling in step 3. The usage of getNftProgress is straightforward:
It will output the status ("queued", "processing", or "completed") and progress percent if available. Once completed, you could retrieve the image (though if you plan to immediately mint, you might skip directly retrieving the raw image and just call mintNft as shown next).
To use the prompt enhancement feature through the SDK, call nft.enhancePrompt():
This will return an object whose data.enhancedPrompt is the improved prompt text (just like the REST response). You can then use that string in generateImage or generateNft calls to potentially get better results. Each call costs ~0.5 credits.
After generating the NFT (and obtaining a collectionId), use the SDK’s nft.mintNft() method to retrieve the minting details and finalize the NFT.
In this snippet, replace collectionId with the ID from the generation step (e.g. result.data.collectionId from generateNft() earlier). The mintNft call requires the same three metadata fields: name, description, symbol. On success, mintRes.data will contain the NFT metadata and possibly a transaction reference, similar to the REST response. You should see the image URL (likely an IPFS link) and the collectionId echoed back. If the SDK/backend initiated an on-chain mint, a transaction hash might be present in the data; if not, you’ll use the ABI to mint manually.
After this, the NFT’s metadata is on-chain or ready to be minted on-chain. If no auto-mint happened, the next step would be using a blockchain library to call the actual smart contract’s mint function, as described below.
You can fetch the NFT contract’s ABI via the SDK as well with nft.abi():
This returns the same JSON ABI that the REST call would. You can use this output in your web3 interactions. For example, with Ethers.js you might do:
Where tokenURI is the metadata URL (e.g. the IPFS link) you got from mintNft(). The exact contract address to use would be provided by ChainGPT’s documentation for each chain (or possibly returned in the mint response or available in the getChains data if a new contract is deployed per chain).
ChainGPT’s NFT Generator API uses a credit system. Each API call deducts credits from your balance, so you should understand the costs:
Standard image generation (ChainGPT models): Using models like velogen, nebula_forge_xl, or VisionaryForge costs 1 credit per image at base resolution. If you use enhancement ("1x" or "2x"), it costs 2 credits per image (effectively +1 credit for the upscale). For example, generating a 1024×1024 image with Nebula Forge XL will cost 1 credit, and if you set enhance: "1x" (HD upscale), it will cost 2 credits.
DALL·E 3 model (Dale3): This third-party model has higher costs. Generating a 1024×1024 image with DALL·E 3 uses about 4.75 credits (standard quality). Using HD upscale on it roughly doubles the cost to ~9.5 credits. For other resolutions (since DALL·E can only output 1024×1024 internally, the service resizes if you ask for non-1024 dimensions), the cost is about 9.5 credits (standard) or ~14.25 credits with HD. In short, DALL·E 3 is significantly more expensive per image.
Prompt Enhancement: Each call to the /nft/enhancePrompt endpoint costs 0.5 credits (since it’s a lighter-weight operation).
Minting calls and others: Calling mint-nft or get-chains does not typically deduct credits (they mostly prepare data or fetch info). The main costs are for image generation and prompt enhancement.
Make sure you have enough credits for the operations you plan to run. For example, a single generateNft with Nebula (no enhance) will use 1 credit, plus a mintNft call (0 credits) – total 1 credit. The same with DALL·E 3 might use ~4.75 credits. If you plan to generate many images or use upscale frequently, ensure your credit balance is topped up accordingly.
Both the REST API and SDK provide error information to help you diagnose issues:
HTTP Status Codes: The REST API returns standard codes. A 200 OK indicates success (image or job created). 400 Bad Request means something was wrong with your request (e.g., missing required field or invalid parameter value). For instance, if you exceed a model’s resolution limits, you might get a 400 with an error message like {"error": "height and width exceed allowed resolution for model velogen"}. 401 Unauthorized means your API key is missing/invalid, or you’ve run out of credits – check that you included the correct Bearer token and have sufficient credits. 429 Too Many Requests indicates you hit a rate limit or tried to use the API with no credits (the service may also treat no-credit attempts as a form of rate limit). 500 Internal Server Error is a rare server-side issue – you can retry after a delay or contact support if it persists. The error responses typically include a JSON body with an "error" message explaining what went wrong.
SDK Exceptions: In the Node.js SDK, errors from the API are thrown as exceptions. Specifically, they will be instances of the Errors.NftError class if they originate from a non-200 API response. You should wrap SDK calls in try/catch. For example:
This way you can distinguish ChainGPT API errors (e.g., invalid input, no credits, etc., conveyed in err.message) from other runtime errors. The SDK’s error message will usually match the JSON error from the HTTP response.
Debugging Tips: Ensure that all required parameters are provided (the SDK’s TypeScript definitions can help catch missing fields). If you get a 401, double-check the API key. If an image generation is failing consistently, try the enhancePrompt to improve your prompt, or reduce the resolution/steps in case it’s hitting limits. On rate limits (429), implement a backoff/retry after the suggested time.
You have now generated AI-driven images and minted an NFT using ChainGPT’s API and SDK. From here, you can integrate these calls into your application’s backend or frontend. For a more in-depth understanding of each endpoint and method, refer to the official ChainGPT AI NFT Generator API Reference and SDK Reference documentation. These resources provide detailed descriptions of all parameters, more example scenarios, and information on advanced features.
For any questions or support, check out the ChainGPT community channels or contact the team as listed on the ChainGPT website. Happy building your NFT application with ChainGPT’s AI NFT Generator!