Toolso.AI
Toolso.AI
All ToolsCategoriesTrendingLatest ToolsPricingBlog
Toolso.AI
Toolso.AI
Toolso.AI
Toolso.AI

Discover the best AI tools to boost your productivity

GitHubGitHubTwitterX (Twitter)YouTubeYouTubeTikTokEmail

Popular Categories

  • AI Writing
  • AI Image
  • AI Video
  • AI Coding
  • More Categories

Explore

  • Latest Tools
  • Popular Tools
  • More Tools
  • Submit Tool
  • Pricing

About

  • About Us
  • Contact
  • Blog
  • Changelog

Legal

  • Cookie Policy
  • Privacy Policy
  • Terms of Service
  • Refund Policy
© 2026 Toolso.AI All Rights Reserved
Limited timeLimited-time offerFeatured Listing24h priority review · No backlink · 30 days featured$29.90then $59.90Price rises to $59.90 after Dec 31Ends in--:--:--Submit now
  1. Home
  2. All Tools
  3. Developer Tools
  4. HiAPI
HiAPI interface preview
HiAPI logo

HiAPI

Premium

Build text and media workflows with a single account. HiAPI connects compatible text clients to model-specific protocols and runs image, video and audio generation as background jobs.

Developer ToolsImage GenerationVideo Generation#Text To Speech#Api#AI Video Generation
Start Free Trial
Featured Tool
Saves
Visits
Views
Pricing
Free Trial
Published
Nov 15, 2026
Domain
hiapi.ai
Community rating

Used this tool? Rate it

Rate this tool
Start Free Trial
Featured Tool

HiAPI Product Information

Start Free Trial
Featured Tool
Tool Information
Saves
Visits
Views
Pricing
Free Trial
Published
Nov 15, 2026
Domain
hiapi.ai
Community rating

Used this tool? Rate it

Rate this tool
Featured Tool

Featured Tools

Related Tools

HomeTop toolsSimilar

What is HiAPI?

HiAPI is a developer-facing gateway for AI models from multiple providers, with image, video, music and text generation under one service. Applications connect through an account key. Text clients speak a supported model protocol, while the asynchronous media API gives the application a task ID and delivers the media when processing finishes.

The AI image editing API includes two GPT Image 2.6 variants with separate routes for generating an image and editing one.

Core features

Text protocols and media jobs

  • Text clients. HiAPI lists OpenAI Chat Completions, OpenAI Responses, Anthropic Messages and Google Interactions for text models. The supported text protocol depends on the model.
  • Background generation. For async media generation, a POST to /v1/tasks returns a task ID, with model-specific input fields inside the common envelope.
  • Output state. Media jobs may be queued, processing or archiving before they return a terminal success or failure. An accepted submission can still fail during processing.

GPT Image 2.6 routes

Flare image-to-image accepts a prompt and 1 to 16 JPEG, PNG or WebP source images. Flare separates five rendering quality tiers from planning complexity.

  • Local edits. The Flare image-to-image route supports either an RGBA mask matching the first source image or a transparent annotated PNG, and can lightly redraw outside the target even with a mask. The annotation acts as a guide, and its marks are excluded from the result.
  • Edited image delivery. Flare image-to-image offers 1K, 2K and 4K and can return PNG, JPEG or WebP, with transparency limited to PNG or WebP. JPEG and WebP also have compression controls.
  • Flare and Sunburst. HiAPI positions Flare for speed and Sunburst for detail consistency at medium or higher quality across repeated edits. The two variants have corresponding text-to-image and image-to-image routes.
  • Images from text. The GPT Image 2.6 text-to-image route accepts a prompt of up to 32,000 characters and includes SVG and GIF output alongside raster formats. GIFs offer 2 to 16 frames with frame dimensions and timing controls, automatic aspect ratio and a total frame imagery limit.

Video and audio models

  • Short video from a still. The Grok Imagine 1.5 image-to-video API reference lists clips of 1 to 15 seconds and accepts a single input image at 1080p. The model page FAQ lists only 480p and 720p.
  • Original Grok route. The original Grok Imagine text-to-video route is documented as having 6 to 30 second clips, with 16 to 30 second clips stitched and potentially showing seams.
  • Mixed video reference materials. Seedance 2.0 accepts combined references within separate caps and generates 4 to 15 second videos. On the standard route, reference images and opening or closing frames share an allowance of nine. Reference video and audio each allow up to three clips and each group has a combined limit of 15 seconds. Seedance 2.0 generates synchronized sound by default and supports audio references.
  • Scripted dialogue. ElevenLabs Text to Dialogue is audio only, caps all lines in a request at 5,000 characters and uses the exposed voices without direct cloning of new voices.

Guide

Running a product image job

  1. Create a backend request. HiAPI documents Bearer authentication and tells production applications to keep the API key on the server. The console shows the complete key only when it is created. A product image job submits its sources and edit prompt to the image-to-image route and returns a task ID to associate with the product record.
  2. Choose how to collect the result. Polling reads GET /v1/tasks/:id, and a configured webhook pushes the result on success or failure. Callbacks send a JSON POST containing the task detail itself, only support final delivery, and accept reachable HTTP or HTTPS URLs.
  3. Handle uncertain submission. An identical request with the same idempotency key replays the original task for that account within about 24 hours, even across API key rotation. An in-flight collision returns HTTP 409 with a five-second retry interval. Reusing the same key with a changed payload returns HTTP 422. Replaying a terminal failure within the window still returns that old task.
  4. Record terminal outcomes. A particular callback can be processed more than once, so the task ID identifies duplicates on the receiver's side. Submit idempotency and webhook deduplication cover separate operations.

Callback security and delivery

Optional callback signing uses a shared secret and HMAC-SHA256 over the timestamp and raw body, and callbacks have no signature header when signing is unset. The headers are X-HiAPI-Timestamp and X-HiAPI-Signature. Verification uses the original bytes before JSON parsing. The verification helper rejects timestamps outside 300 seconds and does not deduplicate deliveries. Because the service accepts both HTTP and HTTPS destinations, transport protection is a separate property of the configured URL.

Webhook delivery treats 2xx as success and retries other responses, timeouts and network errors at one, five and 20 minutes. After the first delivery and three retries, the notification stops and the job keeps its terminal state.

Use cases

  • Product catalog variants. E-commerce and brand teams can replace backgrounds, extend a scene or correct details from existing product images. For a catalog using approved source shots, those edits preserve the workflow around the source asset while creating layout variations.
  • Interface assets. Text-to-image SVG and GIF outputs support icon and lightweight animation workflows.
  • Campaign previews. The provider describes short animated draft variants from existing brand concept images for internal review.
  • Scripted audio production. The dialogue route supports separate voices on scripted lines for product demos and podcast segments.

Who is it for

  • Application teams with several model providers. HiAPI offers a single developer key across model providers and handles the infrastructure and common billing. A single application can combine text services and queued media jobs while managing them as separate request types.
  • Teams building agent integrations. The remote MCP bridge supports image, video, audio and text generation tools from compatible clients without local deployment.
  • Creative teams building internal tools. Teams whose reviews require private media links have a separate hosting requirement because the returned links are public.

Platforms

Browser and direct integration

HiAPI provides a browser-based Playground on its model pages to explore parameters and example outputs. The browser has separate Media Generation and multi-turn Chat playgrounds, with reasoning and usage details on the chat side.

HiAPI lists SDKs for Python 3.8+, Go 1.21+, Java 11+ and Node.js 18.17+, wrapping the asynchronous generation API. An SDK polling timeout is 10 minutes by default and the task can keep running and bill after polling stops. The existing task ID can be queried again without submitting another job.

The official SDK guidance documents Node.js creation-time persistent storage, with post-creation promotion or dashboard management as alternatives for Python, Go and Java. Callers using a plain HTTP client can send the documented task field directly when they need the storage choice at submission.

Agent connections

  • Remote MCP. Remote MCP uses Streamable HTTP and OAuth; the hosted ChatGPT flow describes a connection spending cap lasting 30 days.
  • Model-specific instructions. Model-specific Agent Skills are documented for Claude Code, Codex and OpenClaw. The llms.txt index supplies API documentation context for assistants.

Pricing

Balance and billing

HiAPI advertises no subscriptions and lists 2,000 Credits for $1, with a common prepaid balance for media. New accounts receive 200 trial Credits under the published starter reward update. The balance endpoint reports available and used account credit in USD.

The vendor’s terms make purchased balances non-refundable and say they never expire. Termination disables API keys and balances are non-refundable after termination under HiAPI’s terms. The recharge FAQ lists major credit cards, Alipay and WeChat Pay. HiAPI’s terms separately name Stripe as its processor and say the company does not directly store payment card information.

A submission with insufficient balance receives HTTP 402, creates no task and incurs no charge. The amount reserved at submission is automatically returned when an asynchronous task reaches failed status.

The November 24–30 PT event applies the greater of a new account’s first-top-up reward and the Black Friday bonus, without stacking. This event was advertised as upcoming on November 14, 2026. The event page describes that interaction but gives no year-round rate for the first-top-up reward.

Example rates by billing unit

These are the public USD rates captured on November 14, 2026.

ModelExample settingDisplayed rate
GPT Image 2.6 Flare image-to-image1K, low$0.023 per image plus $0.012 per reference image
GPT Image 2.6 Flare text-to-image1K low raster output$0.023 per image
Grok Imagine 1.5 image-to-video720p$0.0214 per second
Seedance 2.0, default720p without reference video$0.293 per second
Qwen Audio 3.0 TTS FlashSpeech synthesis$0.030 per 1,000 characters
Claude Opus 5.5, defaultInput and output$2.020 per million input tokens and $10.080 per million output tokens

Flare and Sunburst have matching rates on corresponding modes. Text-to-image has separate SVG rates by resolution and a GIF rate for each quality tier. Video pricing can vary with references, sound or processing route. Claude Opus 5.5 also has separate cache prices and an AWS route. The terms say prices may change with prior notice.

HiAPI alternatives

  • fal. fal queues requests by model endpoint path, returns request and result URLs and accepts a callback URL. HiAPI passes the model and its inputs inside a shared task route.
  • Replicate. Replicate supports hosted community models, fine-tuning and bringing a custom model. Replicate says API-created prediction inputs and outputs including files are deleted after an hour. HiAPI offers paid retention for selected outputs.

Limitations

Requests and access

  • Shared rate buckets. Default sliding-window limits are 20 writes and 60 reads per 10 seconds per account, shared across the account’s keys. Rejected requests that reach the rate check count, and a rate-limited request returns HTTP 429 with Retry-After.
  • Task ownership. A media task belongs to its creating account and another account gets a 404.
  • Public media links. Anyone who has the public output URL can download the file and HiAPI does not document a private or signed alternative here. This access rule applies even when the task record is restricted to the creating account.

Storage and deletion

  • Temporary by default. HiAPI’s default temporary media storage is free and expires about seven days after creation. Persistent storage can be selected for a task or promoted per output before expiry, and deleting the file ends future charges without a demotion option. The static tier is reserved for platform-hosted assets.
  • Daily billing and capacity. Persistent storage costs $0.05 per GB per month, billed daily at 00:30 UTC, with free downloads. The default account storage cap is 100 GB, with rejected promotions and silent fallback on a new generation that exceeds the cap. Insufficient balance at creation can silently leave an output in the temporary tier. The task detail reports the effective storage tier.
  • Balance lapse. After three consecutive shortfalls for persistent storage, affected files enter a seven-day grace window and are deleted if it expires. HiAPI documents an email warning and says topping up during the window clears it. A top-up after an output has been deleted cannot restore it. Deleted and grace-expired files cannot be recovered through the storage service.

Content policy and rights

  • Moderation is model-specific. Content checks can apply to prompts and reference assets, with rules and outcomes that vary by model. HiAPI’s terms prohibit illegal, harmful or infringing generated material and allow throttling or suspension for abnormal usage.
  • People and likenesses. HiAPI lists a P-Video 2 Pro Replace route that substitutes people within an existing scene and can preserve the original audio. HiAPI says content filtering is always on for P-Video 2 Pro Replace and policy violations fail with a full refund. The public P-Video 2 Pro Replace model listing leaves the rights to the likenesses shown in its examples unspecified. HiAPI’s Google Virtual Try-On FAQ conditions real-person photos on the right to use the person’s likeness and says the examples show generated people. The Google Virtual Try-On FAQ says a safety-filtered result may have no image and the request still bills as one image.
  • Output ownership. HiAPI makes ownership of API-generated outputs subject to the relevant model provider’s terms. The browser image generator also describes commercial use as dependent on provider terms and rights to prompts, source material and outputs.

Privacy and retention

The policy lists account details, call logs, model usage, request logs and technical data among what HiAPI collects, and lists account operations, billing and platform improvement among purposes. The policy says HiAPI does not sell personal information and names Stripe, legal requests and consent in its sharing provisions.

API call logs are retained for 90 days by default. The policy keeps personal information for the account duration and says deleted accounts’ data is deleted or anonymized in a reasonable time unless longer retention is legally required.

The policy describes rights to access, correction, deletion, portability and objection where applicable data protection law provides them. The privacy policy says the service is not intended for users under 16. The public privacy policy leaves model-training use of prompts, uploaded source assets and generated outputs unspecified.

FAQ

Q1. Can a single model name be reused across text and image APIs?

Text model IDs are used on supported text protocols. Media tasks specify their own routes, such as Flare image-to-image, with the model’s media inputs.

Q2. Is a callback required for a small integration?

Polling retrieves the task by ID. A task remains queryable after a missed callback.

Q3. Does a successful media task give me a private link?

Anyone with the link can retrieve the file. The paid tier changes retention and billing.

Q4. Does an empty result always restore the reserved credits?

Terminal failure restores the reservation. Google Virtual Try-On can bill for a safety-filtered result that returns no image.

Know a Similar Tool?
If you know other great AI tools, feel free to submit them to us