Every model, one place.
The foundation models behind LetzAI — image, video, upscaling and text — picked by benchmark and available through one credit balance and one API.
MAI Image 2.7
Microsoft · Nov 2026Microsoft's flagship MAI Image 2.7 — frontier photorealism, typography, and fine-grained localized editing across successive rounds, with native transparent backgrounds and up to 5 reference images at 1K or 2K.
MAI Image 2.7 Flash
Microsoft · Nov 2026Microsoft's fast, lower-cost MAI Image 2.7 variant — rapid image generation and localized editing with up to 5 reference images and native transparent backgrounds at 1K or 2K.
Kling V4
Kuaishou · Nov 2026Kuaishou's Kling 4.0 with native audio, clips from 3 to 30 seconds at 720p or 1080p, automatic shot-splitting from the prompt, and first and last frame control.
MiniMax H3 Image Highspeed
MiniMax · Nov 2026MiniMax's H3 Image in its Highspeed variant — detailed images with precise instruction following across photoreal scenes, illustration and design, at 1K, 2K or 4K. Edits transform and combine up to 16 reference images.
Gemini Omni 1.2 Flash
Google · Nov 2026Google's multimodal video model with native audio, up to 4K and 15 seconds. Handles text-to-video, image-to-video (first and last frame), reference-to-video, and prompt-driven video editing from a single endpoint.
Seedream 5.5
ByteDance · Nov 2026ByteDance's successor to Seedream 5 Pro: stable multi-round editing, native transparent (RGBA) output, sharper small and multilingual text, up to 2K.
P-Image Ideogram
Pruna AI · Nov 2026Pruna AI's fast image model built with Ideogram on Ideogram 4.0 — strong photorealism, precise typography and structured JSON prompts at 1K or 2K in seconds. Edits run on Pruna P-Image-Edit with up to 5 images.
GPT Image 2.6
OpenAI · Nov 2026OpenAI's GPT Image 2.6. Stronger instruction-following than 2.0; different aesthetic. Precise mask-based editing up to 4K.
GPT Image 2.6 Upscaler
OpenAI · Nov 2026AI upscaling via OpenAI GPT Image 2.6 edit with intelligent detail enhancement up to 4K resolution.
Flux 3 Max
Black Forest Labs · Oct 2026Black Forest Labs' Flux 3 Max — fal's post-trained Flux 3 Video with stronger prompt adherence and better visual quality at lower latency. Text-to-video, image-to-video (start frame) and first+last frame keyframes, up to 20s at HD or Full HD with synchronized audio. No Omni / multimodal reference stack.
Flux 3 Image
Black Forest Labs · Oct 2026Black Forest Labs' newest image model — detailed, well-composed images in native 2K and 4K with precise layout control, multi-reference editing (up to 10 images), and improved text rendering.
Nano Banana 2.1
Google · Oct 2026Google's Nano Banana 2.1 image model — multimodal generation and editing with up to 14 reference images, up to 4K resolution and Google Search grounding.
Seedance 2.5 Draft Mode
ByteDance · Oct 2026Seedance 2.5 Draft Mode: fast, low-cost 480p takes of your prompt and references. Iterate on drafts, then upscale only the take you want to the final 1080p video, within 7 days. Text, image, first+last frame and reference-to-video, up to 30 seconds with synchronized audio.
Seedance 2.5 Enterprise Draft Mode
ByteDance · Oct 2026Seedance 2.5 Draft Mode for Enterprise: fast, low-cost 480p takes of your prompt and references. Iterate on drafts, then upscale only the take you want to the final 1080p video, within 7 days. Text, image, first+last frame and reference-to-video, up to 30 seconds with synchronized audio.
Nano Banana 2.1 Upscaler
Google · Oct 2026AI upscaling using Google Nano Banana 2.1 with intelligent detail enhancement up to 4K resolution.
Seedance 2.5 Draft Upscale
ByteDance · Oct 2026Renders a Seedance 2.5 draft as the final 1080p video, with the draft's own prompt, references, length, framing, seed and audio. Only for drafts, within 7 days of the draft.
ElevenLabs v4
ElevenLabs · Oct 2026ElevenLabs' most emotive voice model: 90+ languages, emotion and pacing with audio tags like [whispers].
Gemini 3.8 Flash TTS
Google · Sep 2026Natural Gemini voices that follow a spoken direction.
GPT-4o mini TTS
OpenAI · Sep 2026Fast, low-cost OpenAI speech with steerable delivery.
Seed Audio 1.0
ByteDance · Sep 2026ByteDance's audio model: speech from a described voice, and sound effects, from one prompt.
Eleven Music
ElevenLabs · Sep 2026Full songs with vocals or instrumentals, up to five minutes.
Lyria 3 Clip
Google · Sep 2026Google's music model: a 30-second clip from a description.
ElevenLabs Sound Effects
ElevenLabs · Sep 2026Sound effects and ambiences from a description, up to 30 seconds.
MMAudio v2
fal · Sep 2026Low-cost sound effects from a description.
MiniMax H3 Max
MiniMax · Aug 2026Extremely fast MiniMax video — great for quick tests. Fal's post-trained H3 variant with stronger prompt adherence and aesthetics; text-to-video, image-to-video (optional last frame), and multimodal reference-to-video (up to 9 images, 3 videos, 3 audio; 12 files total), 5–15s at 480p or 768p.
Seedance 2.5 Video Upscale
ByteDance · Aug 2026Upscale a video to 1080p or 4K via Seedance 2.5 reference-to-video. Max 30s.
Seedance 2.5 Enterprise Video Upscale
ByteDance · Aug 2026Org-scoped Seedance 2.5 Studio video upscale to 1080p or 4K. Max 30s.
BytePlus Video Enhancer
ByteDance · Aug 2026Cinema-grade BytePlus MediaKit enhancement (Professional): 30+ algorithms, AI super-res, defect repair, and color. Keeps the original footage (does not regenerate).
Seedance 2.0 Video Upscale
ByteDance · Aug 2026Upscale a video to 1080p or 4K via Seedance 2.0 reference-to-video. Max 15s.
Seedance 2.0 Enterprise Video Upscale
ByteDance · Aug 2026Org-scoped Seedance 2.0 Studio video upscale to 1080p or 4K. Max 15s.
Grok Imagine 2.0
xAI · Aug 2026xAI's newest image model — precise instruction following, sharp typography, and iterative editing with up to 5 reference images. $0.04/image; 1K and 2K.
Seedance 2.5 Enterprise
ByteDance · Aug 2026ByteDance's professional Seedance 2.5 Studio variant with private asset library support. Automatically uploads reference images to BytePlus virtual portrait library for enhanced character consistency. Up to 30s clips with up to 50 multimodal references (30 images / 10 videos / 10 audio) and synchronized audio. 480p, 720p, 1080p, and 4K.
Flux 3 Video
Black Forest Labs · Aug 2026Black Forest Labs' Flux 3 Video — generate and animate video up to 20s at HD or Full HD with synchronized audio. Supports text-to-video, image-to-video (start frame), and first+last frame keyframes. No Omni / multimodal reference stack.
MiniMax H3
MiniMax · Jul 2026MiniMax's frontier video model at 2K. Smart-routes text-to-video, image-to-video (optional last frame), and multimodal reference-to-video (reference images, one video, one audio) while keeping subjects consistent.
Seedream 5 Pro
ByteDance · Jul 2026ByteDance's latest Seedream model with Chain-of-Thought reasoning, enhanced prompt following, reference consistency, and up to 2K resolution.
Seedance 2.5
ByteDance · Jul 2026ByteDance's next-generation Seedance with up to 30-second native clips, up to 50 multimodal references (30 images / 10 videos / 10 audio), and synchronized audio. Supports text, image, first+last frame, and reference-to-video. 480p, 720p, 1080p, and 4K.
GPT Image 2 Upscaler
OpenAI · Jul 2026AI upscaling via OpenAI GPT Image 2 edit with intelligent detail enhancement up to 4K resolution.
Nano Banana 2 Lite
Google · Jun 2026Google's fastest and cheapest Gemini image model — optimized for high-volume generation where speed and cost matter most. Best for single-prompt text-to-image; not optimized for multiple reference inputs or multi-turn editing.
Nano Banana 2 Upscaler
Google · Jun 2026Faster AI upscaling using Google Gemini 3.1 Flash with intelligent detail enhancement up to 4K resolution.
Pruna Upscaler
Pruna AI · May 2026Fast AI upscaling via Pruna P-Image-Upscale with optional detail and realism enhancement up to 128 megapixels.
Seedance 2.0 Enterprise
ByteDance · May 2026ByteDance's professional Seedance 2.0 Studio variant with private asset library support. Supports up to 4K (10-bit color), with text-to-video, image-to-video, and multimodal reference-to-video (up to 9 reference images, 3 reference videos, 3 reference audios) and synchronized audio.
GPT Image 2
OpenAI · Apr 2026OpenAI's GPT Image 2 — the look people actually want. Strong typography, fine detail, and precise mask-based editing up to 4K.
Seedance 2.0
ByteDance · Apr 2026ByteDance's most advanced video model with cinematic output, native audio, real-world physics, and director-level camera control. Supports up to 4K (10-bit color). Supports text, image, audio, and video reference inputs (up to 9 reference images, 3 reference videos, 3 reference audios) with first+last frame control.
WAN 2.7 Image Pro
Alibaba · Apr 2026Alibaba's professional image generation model supporting text-to-image, image editing, and multi-reference generation with up to 4K high-definition output. Supports thinking mode for improved quality.
Nano Banana Pro
Google · Nov 2025Google's state-of-the-art image generation model with improved text rendering, multi-turn editing, and professional-grade controls over lighting, camera, and composition.
Nano Banana Pro Upscaler
Google · Nov 2025AI-powered upscaling using Google Gemini 3 Pro with intelligent detail enhancement up to 4K resolution.
Tested by creative professionals.
We run every candidate through our own benchmark suite and keep only the models that hold up in real creative work.
See LetzBenchOne credit balance for all of them.
Every model draws from the same credits. Compare per-generation costs and pick a plan.
See credit costs
Every model, one API.
Integrate LetzAI into your apps, workflows and agents with the same REST API the product runs on.
- 01
Get your API key
Subscribe to any paid plan, then find your key on the subscription page.
- 02
Generate an image
Send a POST request to create an image with any model.
POST https://api.letz.ai/images - 03
Retrieve your result
Poll the image endpoint with the returned ID until it's ready.
GET https://api.letz.ai/images/:id
Inference partners.
Model availability and pricing may change. Check the pricing page for current rates.
Bring your own keys.
Save an API key from a provider and the models it runs are billed to your provider account instead of your credits. LetzAI keeps 1 credit per job. Included with Pro, Ultra and Enterprise.
- 01
Your key goes first
If you have a key for a provider that runs the model, the job runs there. The model's own host beats fal and inference.sh: Seedream 5 uses your BytePlus key before your fal key.
- 02
Otherwise, our account
Without a matching key, the job runs on LetzAI's account at the model's main provider and is paid in credits.
- 03
A backup when it's busy
When a provider is rate-limited, some models switch to a backup, like Nano Banana Pro moving from Google to fal. A job on your key only switches to a provider you also have a key for.
- 04
fal and inference.sh run more
Their keys also run models we normally host elsewhere, like GPT Image 2 and Seedream 5. Some resolutions or durations can be narrower there.
Which key runs what
ElevenLabs
3 modelsEleven Music, ElevenLabs Sound Effects, ElevenLabs v4
Google
7 modelsGemini 3.8 Flash TTS, Gemini Omni 1.2 Flash, Lyria 3 Clip, Nano Banana 2 Lite, Nano Banana 2.1, Nano Banana 2.1 Upscaler, Nano Banana Pro
OpenAI
5 modelsGPT Image 2, GPT Image 2 Upscaler, GPT Image 2.6, GPT Image 2.6 Upscaler, GPT-4o mini TTS
xAI
1 modelGrok Imagine 2.0
BytePlus
5 modelsBytePlus Video Enhancer, Seedance 2.0, Seedance 2.5, Seedream 5 Pro, Seedream 5.5
fal
34 modelsBytePlus Video Enhancer, Eleven Music, ElevenLabs Sound Effects, ElevenLabs v4, Flux 3 Image, Flux 3 Max, Flux 3 Video, Gemini 3.8 Flash TTS, Gemini Omni 1.2 Flash, GPT Image 2, GPT Image 2.6, Grok Imagine 2.0, Lyria 3 Clip, MAI Image 2.7, MAI Image 2.7 Flash, MiniMax H3, MiniMax H3 Image Highspeed, MiniMax H3 Max, MMAudio v2, Nano Banana 2 Lite, Nano Banana 2 Upscaler, Nano Banana 2.1, Nano Banana 2.1 Upscaler, Nano Banana Pro, Nano Banana Pro Upscaler, Seed Audio 1.0, Seedance 2.0, Seedance 2.0 Video Upscale, Seedance 2.5, Seedance 2.5 Enterprise, Seedance 2.5 Enterprise Video Upscale, Seedance 2.5 Video Upscale, Seedream 5 Pro, WAN 2.7 Image Pro
inference.sh
34 modelsEleven Music, ElevenLabs Sound Effects, ElevenLabs v4, Flux 3 Video, GPT Image 2, GPT Image 2 Upscaler, GPT Image 2.6, GPT Image 2.6 Upscaler, Grok Imagine 2.0, Kling V4, MiniMax H3, MiniMax H3 Max, MMAudio v2, Nano Banana 2 Lite, Nano Banana 2 Upscaler, Nano Banana 2.1, Nano Banana 2.1 Upscaler, Nano Banana Pro, Nano Banana Pro Upscaler, P-Image Ideogram, Pruna Upscaler, Seedance 2.0, Seedance 2.0 Enterprise, Seedance 2.0 Enterprise Video Upscale, Seedance 2.0 Video Upscale, Seedance 2.5, Seedance 2.5 Draft Mode, Seedance 2.5 Draft Upscale, Seedance 2.5 Enterprise, Seedance 2.5 Enterprise Draft Mode, Seedance 2.5 Enterprise Video Upscale, Seedance 2.5 Video Upscale, Seedream 5 Pro, WAN 2.7 Image Pro
Anthropic, OpenAI, Google and xAI keys also run the chat agent.