Updated
AI models and media

Image & Video Generation

Generate images or video inside your product through an API, without running GPUs yourself.

For your framework or language:Next.js
Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗

Trends

Gemini API gained the most ground in package downloads: 4.9× against the median npm package over a year.

Public counts each month: pull requests on GitHub, package downloads, stars, Product Hunt followers. They show direction, not market share.

ToolDownloadsStarsProduct Hunt
OpenAI API37.4M/wk+60% vs npm—34.5k followers
Gemini API22.2M/wk4.9× vs npm—7.7k followers
fal1.4M/wk3.8× vs npm——
Diffusers1.2M/wk★ 34.7k—
Replicate507.8k/wk−13% vs npm—66 followers
Runway50.4k/wk+117% vs npm——
Black Forest Labs21.1k/wk★ 2.8k—
Luma12.8k/wk+11% vs npm—661 followers
ComfyUI—★ 138k—

The real choice

Model marketplace, the model maker's own API, or open models you run. Marketplaces let you switch between many open and closed models behind one account and try new ones the week they ship. A first-party API gives you one model family directly from its maker, with fewer choices but no middleman. Running open models yourself gives full control over pipelines and fine-tunes, but you provide and manage the GPUs.

What makers chose68 live products, each with a source. Not market share.

Pick by situation

Tap the ones that are you — the tools that fit light up below.

Compare the tools

Grouped by the side of the choice they answer, not ranked. Open a row for when to use it, the trade-off and what makers say.

ToolBest for
Model marketplace2
falapiLow-latency image and video generation across many current models, billed per output.20+17 open source1.4M/wk3.8× vs npm

Use it whenGeneration speed is part of your product experience.

Trade-offHosted only; you depend on fal adding and keeping the models you use.

In v0

Used by Fuser, Keet, Magic Hour, PicMotionAI, BetterPic and 15 more · in 17 open-source projects.

FAL has been a game-changer for inference. Plug-and-play API for top-tier AI models without worrying about infrastructure. It lets me ship fast without dealing with GPUs or weird SDKs.
PicMotionAI, the makerSep 2026 ↗
ReplicateapiA wide catalog of community and open models, plus deploying your own fine-tuned model behind an API.10+39 open source507.8k/wk−13% vs npm

Use it whenYou need a niche model or want to host a custom one.

Trade-offBilled per second of compute, and cold starts on less-used models add latency.

llms.txtIn Lovable

Used by Articos, BetterPic, BrandingStudio.ai®, Macaly, Reloop and 5 more · in 39 open-source projects.

Replicate lets us run AI models without worrying about infrastructure. It’s perfect for testing crowd prediction models without spinning up full cloud stacks.
CrowdSynthetic, the makerSep 2026 ↗
First-party model API5
RunwayapiText-to-video, image-to-video and video-to-video generation through Runway's developer API.950.4k/wk+117% vs npm

Use it whenVideo generation is the product, not a side feature.

Trade-offPrepaid credits per generation, and video is slow and costly to iterate on.

Used by Fuser, Bbc, Figma, Havas, Musixmatch and 4 more.

Stability AIapi · self-hostableStable Diffusion models from their maker, with open weights you can later run yourself.3

Use it whenYou want an API now and the option to self-host or fine-tune later.

Trade-offCredit-based pricing, and self-hosting the open weights has its own license terms to check.

Used by Writesonic, AI Game Master- Dungeon RPG, ImageKit.

Excellent Quality at Unbeatable Prices! The quality here surpasses every competitor we've tried, and the pricing is incredibly reasonable. They’ve truly nailed the perfect balance of value and performance. Highly recommended!
AI Game Master- Dungeon RPG, the makerSep 2026 ↗
Black Forest Labsapi · self-hostableFLUX image models from their maker, through its API or, for some versions, as open weights you can run.No evidence yet21.1k/wk

Use it whenYou want FLUX with the newest versions first, or a path to self-hosting later.

Trade-offOpen-weight licenses differ by model, and some forbid commercial use without a paid license.

No maker's product we track shows it for this yet.

Gemini APIapi · free tierGoogle's image models for generation and editing, and Veo for video, on the same key as Gemini's text models.No evidence yet22.2M/wk4.9× vs npm

Use it whenYou're on the Gemini API already, or need image editing that follows instructions closely.

Trade-offOne vendor's model families only, and video generation is priced per second, so costs rise quickly.

In Lovable, Replit

No maker's product we track shows it for this yet.

OpenAI APIapiApps already on the OpenAI API that want image generation and editing with strong prompt following, on the same key.No evidence yet37.4M/wk+60% vs npm

Use it whenYou're on OpenAI already and want images without another vendor.

Trade-offOne model family only, with less control over style than open models you can fine-tune.

llms.txtIn Lovable, Replit, Rork

No maker's product we track shows it for this yet.

Open models you run2
ComfyUIapp · open sourceBuilding multi-step image and video pipelines — models, LoRAs, control inputs, upscalers — as node graphs you run on your own GPU or on Comfy Cloud.1★ 138k

Use it whenYour output needs a specific chain of models and steps that no single API offers.

Trade-offYou run and maintain GPU servers and custom nodes; turning a workflow into a production API is work of its own.

Used by Xindi Zhang.

Diffuserslibrary · open sourceRunning and fine-tuning open image and video diffusion models from Python code.44 open source1.2M/wk

Use it whenYou write the pipeline in code, fine-tune on your own images, or serve a model on rented GPUs.

Trade-offYou manage GPUs, memory and model licenses yourself.

No maker's product we track shows it for this yet · in 44 open-source projects.

Who switches to what

Public pull requests on GitHub since Oct 2024 whose title says "X to Y" — real code changes, by developers in general rather than makers only. Pick a flow to see its pull requests.

OpenAI API → Gemini API178+ PRs

Before you choose

How to approach it

Pick the model by generating the same 20 prompts from your real use case on a few candidates, then pick the platform that hosts it. Generations are slow and cost per output, so run them in a background job and store results in your own object storage. Set per-user limits from day one; image generation is an easy way to get a surprise bill.

Common mistakes
  • Generating inside the web request, so users stare at a spinner until the function times out.
  • Serving the provider's temporary output URLs to users instead of copying results to your own storage, and finding the images gone a day later.
  • Launching free generations with no per-user limit, then paying for someone's script that ran all night.

Other options

Real choices most makers here won't need to weigh.

Decided alongside

What the 68 makers' products here chose for their other decisions.