Blog
AI Image & Video API Guides
Hands-on model tests, production tutorials, prompt techniques, and current API pricing for generative media teams.
Seedream 5.0 Lite Image-to-Image Prompt Recipes: 3 Copy-Paste Edits With Real Before/After Results
Three copy-paste edit recipes for seedream-5.0-lite/image-to-image, each shown with the real before/after pair it produced — all first takes. Covers the three editing jobs that matter: a precise local edit, a multi-reference subject-into-scene composite, and a text replacement on a sign. Includes the task-API call with the required image_urls/aspect_ratio/resolution fields, the 2K/4K rule, and a five-rule edit-prompt pattern.
Wan 2.7 Text-to-Video Prompt Recipes: 3 Copy-Paste Prompts With Real Outputs
Three copy-paste text-to-video prompt recipes for wan2.7-video/text-to-video, each shown with the real, first-take clip it produced — no re-rolls. Covers the three core t2v prompt jobs: building a photoreal scene, locking a stylized medium, and directing a camera move with a reveal. Includes the validated task API request shape, native-audio prompting, and per-second pricing — all three clips cost $1.84 total.
Generating Product Images at Exact Dimensions via API: Aspect Ratios and Marketplace Specs
hiapi image models take aspect_ratio plus a 1K/2K/4K resolution class -- never pixel sizes -- so exact marketplace dimensions come from generating at the matching ratio and doing one local LANCZOS downscale. gpt-image-2 covers 16 aspect ratios (including social-only 4:5 and 2:3); wan2.7-image adds 8:1/1:8 ultra-wide banner strips and is a solid 4K pick. Includes a marketplace spec cheat sheet (Amazon, Shopify, Instagram, Pinterest, Etsy) and a batch Python script that creates all tasks up front, polls, and saves each channel at exact pixels.
Seedance 2.0 One-Shot Video: Anatomy of a 4K POV Flythrough Prompt
The clip in this post is one continuous 15-second POV shot from a single Seedance 2.0 prompt — no cuts, no keyframes, and the audio was generated with it. The prompt does six jobs: lock the camera, ban its carrier, stack the corridor with near-misses, script the ending, fix the grade, and score the sound. Seedance 2.0 renders native 4K (3840×2160); iterate at 720p ($0.293/s) and re-run the keeper at 4k ($1.486/s, July 2026 pricing). Full prompt and a copy-paste /v1/tasks request included.
Kling AI Short Video via API: An End-to-End kling-3.0-omni Text-to-Video Workflow
End-to-end workflow for short-form vertical video with kling-3.0-omni/text-to-video on the hiapi API: submit an async task, poll, download the .mp4. Four real first-take clips embedded with their exact prompts: a latte-art hook, a native-audio wok scene, a 10s three-shot story, and a batch variant. Native audio requires sound: true plus an explicit soundscape in the prompt; silent clips ship with no audio track at all (verified with ffprobe). Multi-shot prompts work by numbering shots and repeating one visual anchor; budget about three seconds per shot. 720p is $0.10/s ($0.143/s with audio) — a 30-clip month of 5-second hooks is $15; everything in this guide cost $3.15.
Wan 2.7 Image-to-Video Prompt Recipes: 3 Copy-Paste Prompts With Real Outputs
Three copy-paste image-to-video prompt recipes for wan2.7-video/image-to-video, each shown with the input frame it started from and the real, first-take clip it produced. Covers the three core i2v jobs — product motion with a locked camera, portrait micro-performance, and explicit camera movement — plus the task API request shape and schema quirks. All 14 seconds of finished 720P video in this post cost $1.40 at the standard $0.10/second tier, with every price checked against the live pricing API.
Seedream 5.0 Lite Prompt Recipes: 6 Copy-Paste Prompts With Real Outputs
Six copy-paste prompt recipes for seedream-5.0-lite/text-to-image, each shown with the real, first-take output it produced. Covers the model's three edges: multilingual text rendering (16/16 strings correct), web-knowledge entity grounding, and strict instruction following. Includes the task-API call, the 2K/4K resolution gotcha, and a five-rule pattern cheatsheet for writing your own prompts.
How to Use the MiniMax Music API (minimax-music-1.5): curl, Python, and a Working Request
minimax-music-1.5 generates full songs from a style prompt plus lyrics via hiapi's async /v1/tasks endpoint. The verified input schema takes exactly four fields: prompt (10-300 chars), lyrics (10-600 chars), and optional sample_rate and bitrate enums. Copy-paste curl and Python examples cover create, poll, download, plus production callbacks with when:'final'.
Short-Form Video with the happyhorse-1.1 Text-to-Video API: an End-to-End Workflow
An end-to-end workflow for producing TikTok/Reels/Shorts clips with happyhorse-1.1 text-to-video via hiapi: prompt patterns per short-form genre, the async task lifecycle, and a batch script that renders a week of clips in one run. Three real 9:16 clips with native audio are embedded — lifestyle B-roll, a product-drop teaser, and an oddly-satisfying macro — each next to the exact API request that generated it, plus the per-platform cost math.
Using seedream-5.0-lite/image-to-image for E-Commerce Product Images via the hiapi API
An anchor-and-edit e-commerce workflow with seedream-5.0-lite/image-to-image on hiapi: generate one canonical product shot, then batch background restages, precise hardware fixes, and colorway series from the same reference — every image $0.035, with runnable curl and Python task-API code.
Build with the models you just read about.
Compare image, video, and audio models, then use the same API workflow in production.