Independent technology guide
The Original AI Models Explained
A practical field guide to the companies that train foundation models, what those models cost, what they are designed to do, and what people can build with them.
Category 02
Original image, video, voice & music models
Creative model pricing is often based on images, seconds, minutes, credits or monthly subscriptions rather than text tokens.
OpenAI
Generate and edit images, create video scenes and commercials, run voice agents, transcribe and translate speech.
Google DeepMind
Create product images, advertising videos with audio, music, speech recognition and dubbing.
SpaceXAI
Generate and edit images or videos and build conversational voice experiences.
Meta
Create social images and edits; research systems also cover video and music or audio generation.
Amazon
Create advertising images, controlled edits, generated video and real-time voice assistants.
Black Forest Labs
Create photorealistic images, product mockups, inpainting, outpainting, background replacement and virtual try-on.
Stability AI
Self-host image generation and editing; create upscaled visuals, 3D assets and audio or music.
Midjourney
Create artistic images, concept art, fashion inspiration, promotional visuals and short videos.
Adobe
Create commercial marketing graphics, Photoshop edits, vector art, video, sound effects and audio translations.
Runway
Create commercials and cinematic video, image-to-video, video edits, consistent characters and lip-sync.
Kuaishou
Create realistic text-to-video and image-to-video scenes with detailed camera motion.
ByteDance
Generate cinematic video, multimodal creative work, images and social-media advertising.
Luma AI
Generate video and images, animate still pictures, add cinematic motion and capture 3D scenes.
Pika Labs
Create short social videos, special effects, object transformations and animated images.
Ideogram
Create images with strong typography, including logos, posters, banners and advertising concepts.
Recraft
Create editable SVG artwork, vectors, logos, icons, packaging, print designs and product mockups.
ElevenLabs
Create voiceovers, authorized voice clones, dubbing, transcripts, sound effects, music and phone agents.
Suno
Generate complete songs, lyrics, vocals, instrumentals, stems and custom music variations.
Udio / Uncharted Labs
Generate songs, extend tracks, remix music and create vocals or instrumentals.
Important distinction
Popular AI products are not always original models
An AI product may combine one or more foundation models with search, tools, interfaces and business data.
ChatGPT
OpenAI’s application powered by GPT, image, audio and Sora model families.
Microsoft Copilot
A product that can combine Microsoft systems, OpenAI models, search and tools.
GitHub Copilot
A coding product that can route requests to models from several providers.
Perplexity
An AI search product using proprietary adaptations and external foundation models.
Poe
A platform that gives users access to models and bots from multiple companies.
Canva Magic Studio
A design suite combining Canva systems with multiple partner technologies.
Replit & Lovable
Software-building agents that commonly use outside foundation models.
OpenRouter
An API marketplace and model router—not a foundation-model developer.
How to choose the right AI model
Start with the job, then compare quality, total cost, data policy, speed, deployment choices and licensing. The most famous model is not automatically the best fit.
Comments