Model Kit.
2 · UnderstandEvery model you need, behind one interface. Build on the best model for each job and switch when something better or cheaper arrives. Your product keeps running while the model landscape keeps moving.
104 APIs in this kitAPIs in the Model Kit
104 servicesResume Parser API
Parse resumes and CVs into JSON: contact details, work history, education, and skills. Works with PDF, Word, and scanned CVs in 80+ languages. Free tier.
SharpAPIVerifiedRun Graph Algorithms API
Run BFS, DFS, shortest path, topological sort, minimum spanning tree, and distance matrix calculations on a graph described by vertices and edges.
Basic Sentiment Analysis API
Detect positive, negative or neutral sentiment in text, with an overall score plus per-sentence results and confidence. Simple REST API, JSON output.
SkycraftVerifiedText Analysis & Scoring API
Score text readability with Flesch, Fog, SMOG and more, plus style, skimmability and interestingness scores, word and sentence counts. JSON REST API.
SkycraftVerifiedText Correction & Word Tools API
Correct spelling in text, list synonyms for a word, and singularize or pluralize nouns with one REST call. Clean JSON output, no word lists to maintain.
SkycraftVerifiedText Entity & Date Extraction API
Extract people, places, organizations and dates from free text, with normalized date values and character positions for every match. Fast JSON REST API.
SkycraftVerifiedChat Completion with Qwen3.8 Flash 27B
Generate text with Qwen3.8 Flash 27B. JSON chat completions, model discovery, and usage-based metering for coding and text workflows.
FlonnoGenerate Image Foreground Mask API
The Image Mask API finds the main subject in an image and returns a grayscale mask of it. Use the mask to composite, clean up edges or analyse the subject in your own pipeline. Send the image in image, as a public URL or base64 data, up to 20 MB. Set format to json (default) or png. JSON returns the mask as a data URI with its metadata. PNG returns the mask file. Set quality to best (default) or fast. Set crop to true to trim the mask to the subject, with padding (0 to 1000 pixels) of space around it. Use maxsize to limit the output size. The JSON response includes the width, height, bytes, the subject bbox in the original image and the foregroundratio, the share of the image covered by the subject. If no subject is found, bbox is null. Teams use the Image Mask API to build cutouts for product photos and to place subjects on new backgrounds in their own editor. Others check that an upload has a clear subject, or crop images to it. To get the cut-out subject instead of the mask, use the Remove Image Background with Advanced Controls API. To read text from images, use the Image & PDF OCR to Markdown API.
Callable LabsGenerate Text to Video With LTX 2 Video (RTX 5090)
This service turns a text prompt or a source photo into a short LTX-2 video on a dedicated NVIDIA RTX 5090, and gives you a task ID to track progress. Text to Video — Send a prompt, and optionally seed, model, length, quality, and resolution. Use the POST /v1/video/generate endpoint to start the job. length supports 3s, 5s, or 10s, quality can be fast, balanced, or best, and resolution can be sd or hd. Image to Video — Send an image (base64-encoded) and a prompt describing the motion, and optionally seed, length (3s, 5s, or 10s), quality (fast, balanced, or best), and resolution (sd or hd). Use the POST /v1/video/i2v-ltx2 endpoint to animate a photo while keeping the composition of the source image — a good fit for ad mockups, product demos, and social content. Both endpoints respond with status, message, and taskid so you can store the job and poll later. Poll the shared GET /v1/video/tasks/:taskid endpoint until the job reaches completed or error. The status response includes type, status, taskid, elapseds, and, when generation finishes, an output_url for the final MP4. If something fails, check the error field on the task response. This service is a good fit for ad mockups, concept previews, social clips, and internal creative iteration where a short video draft is enough — whether starting from a text description or animating an existing photo. Notes: Text-to-video and image-to-video with LTX-2 19B (distilled) on a dedicated NVIDIA RTX 5090. Text-to-video generates a ~5-second cinematic clip in ~30-60 seconds; image-to-video generates a ~5-second clip in ~30-60 seconds.
MasterA AIVerifiedGenerate Text to Video With LTX Video (RTX 5090)
This service turns a text prompt or a source photo into a short MP4 video using LTX video generation on a dedicated NVIDIA RTX 5080. Text to Video — Send a prompt, and you can optionally tune cfg, seed, steps, width, height, length, and negativeprompt to control guidance, reproducibility, frame count, and output size. Use the POST /v1/ltx/generate endpoint to start the job. Image to Video — Send an image (base64-encoded) and a prompt describing the motion, and optionally tune quality (fast, balanced, or best), resolution (sd or hd), negativeprompt, and seed. Use the POST /v1/ltx/i2v endpoint to animate a photo while keeping the composition of the source image. Use it to animate a product photo, portrait, or scene image into a short cinematic clip. Both endpoints start a generation task and return a status, message, and taskid. Use the shared GET /v1/tasks/:taskid endpoint to check progress until the job reaches completed or error. When the video is ready, the task status response includes an output_url to the finished MP4. Use this service when you need short motion clips from a script, product concept, storyboard note, or scene description — or when you want to bring a still image to life. It is a fit for prototyping video ideas, creating demo assets, animating product photos, or generating synthetic motion content without a manual editing workflow. Because outputs are task-based, you can kick off generation and poll asynchronously while your app continues other work. That makes it easy to slot into content pipelines, creative tools, or internal review systems where you need a generated clip and a reliable completion signal. Notes: Text-to-video and image-to-video with LTX-Video 13B (distilled) on a dedicated NVIDIA RTX 5080. Text-to-video generates a ~4-second cinematic clip in ~30 seconds; image-to-video generates a ~5-second clip in ~20-40 seconds depending on quality/resolution tier.
MasterA AIVerifiedGenerate Text to Image Job API
GPU Image Generation creates images from text prompts and returns a task you can poll for progress. Use it when you need fast, model-backed image generation without managing your own inference stack. Send a prompt and, optionally, seed, steps, width, and height to the Flux.1-Schnell endpoint or the SDXL-Turbo endpoint. The Turbo variant also accepts negativeprompt so you can steer the output away from unwanted elements. Both generation endpoints respond with status, message, and taskid, which you use with GET /tasks/:taskid to check progress. Polling /tasks/:taskid returns the current status, a human-readable message, the taskid, and, when the task is complete, an outputurl for the generated file. If the task fails, the same status response includes error. This service fits product features like prompt-based asset creation, concept mockups, social content generation, or internal design tooling where developers need an image job API with asynchronous results and reproducible seeds.
MasterA AIVerifiedAI-Generated Image Detection API
Classify a public JPG, PNG, or WebP URL as likely AI-generated, human-made, or uncertain. Returns ai_score, confidence, and a source breakdown.
AnyImageDetectorWhich Model API do you need?
The question you arrived with, and the API that answers it.
What is the Model Kit?
The Model Kit is the model layer of your product: chat, reasoning, summarization and classification through one consistent interface across providers.
You choose the right model per task, compare cost and quality, and move fast when the market shifts, without rebuilding integrations each time.
How AI products use it
- Send chat to one provider and summaries to another, through the same interface.
- Switch providers when price or quality changes, without rewriting code.
- Add new models as they launch while your integration stays the same.
Why not build it in-house?
Each capability in the Model Kit looks simple to build. The hard part is keeping it accurate, available and up to date once your product depends on it.
- Provider integrationsEach LLM provider has its own SDK, auth, error codes and rate limits. Every one you add is another integration to keep current.
- Model churnNew models launch and old ones are retired every few months. Each switch means re-testing prompts, outputs and costs across your product.
- Outages and rate limitsWhen a provider slows down or goes down, your product does too, unless you have built retries, fallbacks and quota handling yourself.
- Cost trackingToken prices differ per model and change often. Knowing what each feature costs you needs metering you have to build and maintain.
Do you run an API that belongs in the Model Kit?
Publish it on ApyHub and put it in front of the teams and AI agents building with the Model Kit. Your API joins the catalog under one subscription, becomes callable through MCP, and earns from every call.
Questions about the Model Kit
What is the ApyHub Model Kit?
The Model Kit is the model layer of your product: chat, reasoning, summarization and classification through one consistent interface across providers.
What is an LLM API?
An LLM API lets your software send prompts or chat messages to a large language model and receive generated text back. It is how AI products add chat, summaries, classification and agent reasoning.
Is the Model Kit OpenAI-compatible?
Yes. Chat Completion Generation follows the OpenAI chat format, so code written for OpenAI-style chat completions works with minimal changes.
Can I choose the model provider per request?
Yes. You choose the provider on each request, so you can route different tasks to different providers and switch when price or quality changes.
When should I use the advanced chat completion API?
Use Advanced Chat Completion Generation when you need tools, streaming, web search or structured output. Use Simple Chat Prompt for a single prompt and response.
Can I test Model Kit APIs before integrating them?
Yes. Every API has a Try it panel on its service page. Your key is filled in, you send a real request and see the actual response in your browser before you write any code.
Can AI agents use the Model Kit?
Yes. Every endpoint is MCP-ready. Connect ApyHub MCP to Claude, ChatGPT, Cursor, Mistral Le Chat or any MCP-compatible agent, and your agent can discover and call these APIs directly.
How is it priced?
Every kit sits under one subscription, paid in atoms, ApyHub's usage unit. Every call shows its cost, and the free Starter plan needs no card.