What is OpenRouter and why does the Creative category matter?

OpenRouter is a routing layer that lets developers call dozens of AI models through one API with per-token billing. The "Creative" sub-category covers apps that generate images, voice, video, and other media. As of April 2026, it grew roughly 4× in three months, making it the fastest-growing app category on the platform.

Which apps are at the top of the Creative rankings?

As of April 26, 2026: Descript (5.73M weekly requests, AI video and podcast editing), CoffeeCat AI Image Generator (1.73M requests), VidMuse (1.61M, audio-to-video music), Fish Audio (374K, TTS and voice cloning), and novelcrafter (850M tokens, novel writing). Token leaders: Descript at 44.5B, VidMuse at 8.12B.

How much does this cost the apps building on OpenRouter?

It varies by which underlying models they route to, but founders building image and voice features on OpenRouter routinely report $200-2,000 per month in token spend by the time they have a few thousand active users. The per-call meter compounds with every prompt, every retry, every variant.

Is this growth sustainable for the apps?

Probably not at the per-token model. The pattern we keep seeing in founder interviews: app launches, finds product-market-fit, scales user count, then realizes 60-80% of revenue is going to AI infrastructure. The next 12 months will see a wave of these apps either raising prices or moving the media layer off per-token billing.

What is the alternative to per-token billing for image and voice?

Flat-rate API providers that charge a monthly subscription instead of per-call. Gathos is one such provider · $18/month flat covers unlimited image generation and TTS with voice cloning across 600+ languages. The shift from per-token to flat is the same shift Spotify ran on the music industry, just compressed into 18 months instead of a decade.

All posts

Industry · Token EconomicsApril 26, 2026 · 10 min read

Creative AI apps quietly became the highest-token-burning category on OpenRouter in two months

Between January and April 2026, OpenRouter's Creative app category went from roughly 85 billion tokens per week to over 340 billion. We pulled the numbers, talked to founders building on top of it, and mapped out what the cost math actually looks like for the apps at the top of the rankings.

By The Gathos team · API platform for AI agents

Details unavailable.

Keep reading

YouTube · Multilingual

YouTube multi-language audio is the biggest creator opportunity of 2026. Most ignore it.

YouTube's multi-audio feature unlocks subscriber growth that doubles or triples per video. Adoption is still under 4% of channels because the dub workflow used to take 12 hours. Voice cloning collapsed it to 8 minutes.

TTS APIs

Best Text-to-Speech APIs in 2026: A Developer's Comparison

A practitioner's comparison of the six most-used TTS APIs in 2026 — ElevenLabs, OpenAI, Cartesia, Google, AWS Polly, and Gathos. Honest takes on pricing, latency, voice quality, code, and how to pick the right one.

Podcast Growth · Workflows

Podcast growth is now a clipping problem. The economics broke last quarter.

Spotify and Apple drive 60% of podcast subscription, but 100% of new listener discovery in 2026 comes from Reels, Shorts, and TikTok. The clipping bottleneck is the part that finally broke open.

Gathos — AI Image, TTS, and Video APIs for Agents

Gathos is an API platform for AI agents. It ships three creative REST APIs: image generation with pixel-perfect long-text rendering, text-to-speech with zero-shot voice cloning across 600+ languages, and Creator text-to-video with generated audio. Pro is $18 per month; Creator adds video at $45 per month.

What is Gathos?

Gathos replaces the common stack of Nano Banana Pro + ElevenLabs + Midjourney, and Creator reduces the need to stitch separate Veo 3 or Seedance-style video workflows onto that stack. It is designed for shell-capable AI agents including Claude Code, Cursor, Windsurf, Gemini CLI, OpenClaw, Aider, GitHub Copilot, ChatGPT Custom GPTs, and Continue. The image API renders readable long paragraphs, headlines, and UI labels inside generated images — the exact thing that competing models (Nano Banana Pro, Midjourney, DALL·E, ChatGPT image) consistently garble. The TTS API clones any voice from a 5–30 second reference sample, zero-shot, and synthesizes speech in more than 600 languages and accents.

Pricing

Pro plan: $18 / month for image and TTS APIs. Creator plan: $45 / month and adds text-to-video with generated audio. No per-image fees or per-character metering. Cancel anytime. Free trial: 7 days with up to 20 calls per day across the creative APIs (140 calls total), the same 6-hour burst window and concurrency limits as the paid plan. No credit card required. Competitor benchmarks (current as of April 2026): Nano Banana Pro via Gemini API is $0.134/image at 1K–2K resolution and $0.24/image at 4K, so 1,000 images costs $134–$240. ElevenLabs TTS with voice cloning runs $5–$990 per month depending on character volume and tier. Midjourney is $10–$120/month with image quotas. Gathos at $18 flat saves a typical mid-volume team $150–$1,500 per month.

Features

Image Generation API — pixel-perfect long-text rendering, outputs 1024×1024 / 1024×1280 / 864×1536 / 1536×864 PNGs, any visual style.
TTS API — zero-shot voice cloning, 6 preset voices (Josh, Koko, Pixxy, Prof, Rochie, Spraky) plus unlimited custom voices, 600+ languages.
Video Generation API — Creator text-to-video with generated audio, optional Styles, async polling, and MP4 output.
Agent-native REST — Bearer-token authentication, standard JSON request/response, async job polling for images and video.
Pre-built MIT-licensed agent skills — Idea-to-Presentation (.pptx + narrated .mp4), YouTube Video Factory (1920×1080 .mp4 + thumbnail), Script-to-Reel (1080×1920 vertical .mp4). One-line install: curl -sL https://gathos.com/install.sh | bash.
Unlimited calls at one flat price — no credits, no per-image fees, no metered tokens.

Frequently asked questions

Which agents does Gathos work with?: Any shell-capable AI agent, including Claude Code, Cursor, Windsurf, Gemini CLI, OpenClaw, Aider, GitHub Copilot, ChatGPT Custom GPTs, and Continue.
Is the image generator actually better at text?: Yes. Gathos renders long paragraphs and multi-line headlines with correct typography inside images. Nano Banana Pro, Midjourney, and ChatGPT image consistently garble text past a short headline.
How does the trial work?: 7 days, 20 calls per day combined across image and TTS, capped at 140 total. Same 6-hour window and concurrency as the paid plan. No credit card required.
How do I get the agent skills?: Sign in at Sign in and open the Skills tab in your dashboard. Each skill ships with a one-line install command scoped to your API key.