Blog
AI Image & Video API Guides
Hands-on model tests, production tutorials, prompt techniques, and current API pricing for generative media teams.
Qwen Image 3.0 Image-to-Image Prompt Recipes: Copy-Paste Prompts With Real Outputs
Qwen Image 3.0's image-to-image endpoint edits a real photo instead of generating a new one — it keeps your product's exact shape, color, and proportions while changing scene, lighting, or added props. We ran 6 real generations end-to-end: two studio product shots (coffee dripper, sneaker) as starting points, then four image-to-image edits — a sunlit kitchen scene, a styled cup pairing, a golden-hour trail move, and a printed hangtag — all keeping the original product unchanged. Each recipe below is model + input + cost, straight into a copy-pasteable prompt, sitting right next to the actual output image — no guessing what "close" means. i2i costs $0.0398 per image at 1K resolution (base $0.0358 + $0.004 for the one reference photo); text rendered on a tag came out correctly on the first try, but we still checked it by eye.
Qwen Image 3.0 Pro for E-Commerce Product Images
qwen-image-3.0-pro builds a full e-commerce image set: studio hero packshot, lifestyle scene, and a reference-based redraw of the same product into a new setting. The image-to-image endpoint -- not available on base qwen-image-3.0 or qwen-image-2.0-pro -- lets you redraw one clean reference shot into new scenes while keeping the product identity consistent. Verified pricing (2026-09): $0.0497/image at 1K, $0.0993/image at 2K on both endpoints; image-to-image adds $0.004 per reference image. Workflow: submit each asset to POST /v1/tasks, poll GET /v1/tasks/{taskId}, download output[0].url before it expires. For a catalog needing one product across many scenes, generate a single reference shot once and redraw it repeatedly instead of re-prompting from scratch each time.
How to Restyle Images with AI: An Image-to-Image Style Transfer Guide for the hiapi API
Restyling with AI is a regular image-to-image call, not a dedicated endpoint -- the prompt's wording, not a strength slider, controls how much the source image changes. Verified live: neither seedream-5.0-pro/image-to-image nor grok-imagine-quality/image-to-image accepts a strength parameter; both reject it as an unrecognized property. Includes copy-paste curl and stdlib-only Python examples, a two-clause prompt pattern for keeping subject/pose fixed while changing style, and a model comparison for framing control.
First and Last Frame Control in AI Video: A Cross-Model Guide to Precise Motion on hiapi
A cross-model recipe for first-frame and first+last-frame control in AI video on hiapi: which models support it, exact (non-uniform) request fields, and production patterns like callbacks and idempotency.
Qwen Image 3.0 Prompt Recipes: Copy-Paste Prompts With Real Outputs
qwen-image-3.0/text-to-image renders bilingual Chinese/English text accurately enough to use directly, no post-editing required if you proofread the output. The model's aspect_ratio enum is stricter than you'd expect: only 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, and 21:9 are accepted -- 4:5 and 5:4 return a clean, unbilled 400. All 6 recipes below are original prompts we wrote and ran ourselves, each paired with its real, unedited output image. At $0.0358 per image on the task endpoint, running all 6 recipes costs about $0.21 total. Full docs, live pricing, and the model parameter reference are one click away if you want to adapt these prompts to your own product.
Qwen Image 3.0 Pro Image-to-Image Prompts: Copy-Paste Recipes With Real Outputs
qwen-image-3.0-pro/image-to-image takes 1-3 reference images and a text prompt, and holds onto structure (product silhouette, room layout, a face) far better than pure text-to-image re-rolling. Three copy-paste recipes cover the three edits people actually ask for: relighting/restyling a product photo, redecorating an interior without moving the walls, and restyling a portrait while keeping the person recognizable. Every prompt is one we wrote and ran ourselves - real before/after pairs, not stock examples. Pricing is flat and predictable: $0.0497 per image at 1K, $0.0993 at 2K, plus $0.004 per reference image passed in (1-3 allowed). A single-reference 1K edit costs $0.0537. The model keeps geometry and identity anchored to your reference - it's an editor, not a re-imaginer.
qwen-image-3.0-pro/text-to-image Prompt Recipes: Copy-Paste Prompts With Real Outputs
qwen-image-3.0-pro/text-to-image pricing is resolution-tiered: $0.0497/image at 1K, $0.0993/image at 2K. The API takes aspect_ratio + resolution, not raw pixel dimensions. 21:9 aspect ratio is currently locked to 1K resolution only. Six full copy-paste prompts cover icons, product photography, banners, wallpapers, landscapes, and portraits. Every image below is a real, unedited task-API output — no stock photography.
How to Use veo-3.1-lite/text-to-video via the hiapi API: curl, Python, and a Working Request
Working curl and Python examples for calling veo-3.1-lite/text-to-video through hiapi's unified async task API. Covers the exact input schema (prompt, duration, resolution, aspect_ratio, generate_audio, negative_prompt, seed), callbacks, and the two distinct error shapes (400 validation vs 401 permission_denied) you'll hit in production.
How to Use veo-3.1-lite/image-to-video via the hiapi API
A working curl and Python example for the veo-3.1-lite/image-to-video model on hiapi's async task API. Covers the exact input schema, callbacks vs polling, idempotency, and how to handle 401/400 errors in production.
flux-2-klein-9b API: curl & Python Guide
flux-2-klein-9b takes just one required input, prompt, through the same POST /v1/tasks endpoint every hiapi model uses. This guide captures a real request and response: task creation, polling, and the final image URL. The model id must include the /text-to-image endpoint suffix or task creation fails with MODEL_UNAVAILABLE. The input schema is strict — there's no size or quality field, only prompt, aspect_ratio, seed, and output_format.
Build with the models you just read about.
Compare image, video, and audio models, then use the same API workflow in production.