Blog
AI Image & Video API Guides
Hands-on model tests, production tutorials, prompt techniques, and current API pricing for generative media teams.
ElevenLabs Text-to-Dialogue API for E-Commerce Audio: Product Video Voiceovers and Ad Reads
ElevenLabs Text-to-Dialogue turns a scripted conversation into one multi-speaker audio file — assign a different preset voice to each line, no separate TTS calls or manual stitching. Priced at $0.125 per 1,000 characters through the hiapi API, billed on the total dialogue script length, not per line or per speaker. The live request schema only accepts `text` and `voice` per line — despite the model's own description mentioning a "stability" control, the deployed API does not expose that parameter today. Good fit for e-commerce teams producing two-host product demo voiceovers, multi-character ad reads, or promo scripts without booking voice talent per SKU. Runs on hiapi's standard async task flow (submit, poll, download), so it drops into the same pipeline as every other model on the platform.
GPT Image 2.5 vs Nano Banana Pro: Which Needs Less Fixing?
Flare rendered NIGHT SHIFT once; Nano Banana Pro added a second SHIFT below the headline. Flare placed the shoe toe left; Nano Banana Pro placed it right. For these two briefs, Flare is the first model we would try; the test covers Flare only, not every Image 2.5 model.
GPT Image 2 Transparent Background: Generate a PNG Without Code
Choose 1K and Transparent in the GPT Image 2 Playground. Download the PNG and place it on your real design background before delivery. The handheld sample has a warm edge halo, so a transparent file can still need a quick cleanup.
GPT Image 2 vs Nano Banana Pro: Is Nano Worth $0.027 More?
The 1K regular-price gap is $0.027 per image. For these two prompts, the higher Nano Banana Pro rate did not produce a clear reason to choose it. Both poster outputs spelled the requested copy correctly; GPT Image 2 used a left information block, while Nano Banana Pro built a centered poster. GPT Image 2 followed the headphone instruction that both earcups face the viewer. Nano Banana Pro kept the colors and three controls but turned the earcups inward.
MiniMax Music 2.6: Generate Background Music for Short-Form Video and Ads
minimax-music-2.6's Instrumental Mode generates royalty-free background music for short-form video and ads in one API call, with no vocals and no lyrics field required. Pricing is a flat $0.21 per song regardless of length, unlike minimax-music-3's per-second billing — easier to forecast for batch background-music production. Omitting the `lyrics` field is what triggers Instrumental Mode; there's no separate boolean flag in the schema. We ran a real prompt end-to-end and share the exact input and the generated track below.
Using glm-5.3 for E-commerce Copywriting and Support Replies
glm-5.3 is a text-only model on hiapi with no image generation — this guide covers real e-commerce copywriting and support-reply use cases instead. Two live API calls show real token usage: a product listing (89/1146 tokens, $0.0180) and a support reply (87/358 tokens, $0.0059). glm-5.3 has no selectable reasoning_effort parameter, and reasoning-token spend is inconsistent — the simpler copywriting task used far more reasoning tokens than the policy-reasoning support task.
ElevenLabs Music API Alternative: Generating Music with hiapi's minimax-music-3 and lyria-3.5
minimax-music-3 generates full instrumental or vocal tracks up to 5 minutes at 44.1kHz WAV, priced per-second ($0.0026/s) — best when you need longer, more controllable output. lyria-3.5 takes custom lyrics and returns a finished song for a flat $0.09, output as compressed audio — best when you want a quick, lyrics-driven track without managing duration. Both run through the same hiapi task API you'd already use for image or video generation — one key, one auth pattern, no separate vendor account. We generated one instrumental track (minimax-music-3) and one custom-lyrics vocal track (lyria-3.5) from real prompts for this piece — both are linked below as playable files. Neither model exposes a lyrics-to-timestamp export or stem separation today; if your workflow depends on those, check the model pages for current capabilities before committing.
DeepSeek V4 Pro for E-Commerce: Product Copy and Support Replies
DeepSeek V4 Pro has no image endpoint — it's a text/reasoning model, so "product images" routes to copy and support text instead. A full product listing generated from a raw spec sheet cost $0.0088 with zero invented specs. A policy-aware customer support reply that correctly separated return eligibility from who pays shipping cost $0.0040. Reasoning is on by default for this model, even for one-line prompts, unlike most other text models on the platform.
How to Use glm-5.3 via the hiapi API: curl, Python, and a Working Request
glm-5.3 is available on hiapi through the standard OpenAI-compatible /v1/chat/completions endpoint, not the async /v1/tasks flow used for media models. This guide walks through a minimal curl and Python request, then covers the production patterns: streaming via SSE, tool calls, and the reasoning-token budget gotcha that returns null content on low max_tokens. Every request and response shown was verified live against the API while writing this piece, including a truncated-by-reasoning response, a real tool-call round trip, and the 401 error format.
How to Use Claude Sonnet 4.6 via the hiapi API
Claude Sonnet 4.6 is available on hiapi through the standard OpenAI-compatible /v1/chat/completions endpoint, not the async /v1/tasks flow used for media models. This guide walks through a minimal curl and Python request, then covers the production patterns: streaming via SSE, function/tool calls, and opt-in adaptive thinking with effort control. Every request and response shown was verified live against the API while writing this piece, including the 401 error format and a real tool-call round trip.
Build with the models you just read about.
Compare image, video, and audio models, then use the same API workflow in production.