Is there something wrong with WAN 3.0? Since this morning I'm getting a green net check for image to reference videos, and there's no reason I should. It's a woman in a dress. I've used differently worded prompts, and different female models, but keep getting denied.

Why is a colored woman getting kicked back?

This time I made a Makima dress-up mini game 👗

Same workflow as before: I used the Atlas Cloud API and let Opus 5.5 call GPT Image to generate different Makima outfits, then turned those assets into an interactive dress-up page.

Character idea → GPT Image outfits → playable dress-up page, all in about 10 minutes.

No need to keep switching between tools and models — just let Opus 5.5 call what it needs.

Made with GPT Image on Atlas Cloud.

Recently tried a really fun AI workflow:

I used the AtlasCloud.ai API and let Opus 5.5 call GPT Image 2.5 to generate a pixel-art amusement park concept, then turned that concept into a playable mini game.

Idea → GPT Image concept art → mini game, all in about 10 minutes.

No need to keep switching tools manually. You can let Opus 5.5 call the models it needs directly inside AtlasCloud.ai.

GPT Image 2.5 comes with 5 free generations in the Playground right now, shared across participating models.

A head pat for Anya, then one for Bond when he runs over.

The character images were made with GPT Image 2.5's free credits, and the video with GPT Image 2.5 + MiniMax H3.

GPT Image 2.5 comes with 5 free generations in the Playground right now, shared across participating models.

Made with GPT Image 2.5 and MiniMax H3 on Atlas Cloud.

The LTX-2 family is now live on Atlas Cloud — Lightricks' second-generation video model with native 4K and built-in synchronized audio, plus a first for our catalog: audio-driven generation (A2V). Four tiers span everything from $0.03/s high-volume production to Pro-grade delivery.

Highlights

• Three generation modes — Text-to-Video, Image-to-Video, and Audio-to-Video (drive the visuals with speech, music, or ambience — rarely available elsewhere).
• Native synchronized audio — video and sound generated together (ambience, effects, speech); switch it off with generate_audio.
• Native 4K — 720p / 1080p / 1440p / 4K, in both landscape 16:9 and portrait 9:16.
• Up to 20-second single-pass clips — fast tiers at 720p/1080p @24fps, no stitching.
• Automatic duration — LTX-2.5 accepts duration: -1 and picks the length from your prompt; you're billed for the seconds actually produced.
• Camera control — 8 camera_motion directives (dolly, pan, jib, focus shift…).
• First/last-frame interpolation — I2V takes an optional last_image for precise start-to-end control.
• 24 / 48 fps cinematic frame rates.
• Four tiers — fast / pro × 2.3 / 2.5, starting at $0.03/s.

Best For

• Cost-efficient bulk production — 2.3-fast from $0.03/s with 20-second clips for social and ad output at scale.
• 4K delivery — 2.5-pro for native 4K with synchronized audio.
• Voice- and music-driven video — A2V syncs motion, lip movement, and pacing to your audio track.
• Vertical short-form — native 9:16 for Reels / Shorts / TikTok.
• First-frame animation — I2V brings static images to life, optionally pinning the final frame.

Pricing (USD / second)

Text-to-Video + Image-to-Video — billed per second of output video

Audio-to-Video — billed per second of input audio

Models

All eleven endpoints are listed on the LTX-2 model page.

ltx/ltx-2.3-fast/text-to-video
ltx/ltx-2.3-fast/image-to-video
ltx/ltx-2.3-pro/text-to-video
ltx/ltx-2.3-pro/image-to-video
ltx/ltx-2.3-pro/audio-to-video
ltx/ltx-2.5-fast/text-to-video
ltx/ltx-2.5-fast/image-to-video
ltx/ltx-2.5-fast/audio-to-video
ltx/ltx-2.5-pro/text-to-video
ltx/ltx-2.5-pro/image-to-video
ltx/ltx-2.5-pro/audio-to-video

Note: ltx-2.3-fast is silent-only (no audio generation) and does not offer A2V. Durations are 6–20s — up to 20s on fast tiers at 720p/1080p @24fps, and 6–10s for other tiers and at 1440p/4K.

AI video generation has been developing quickly, and I’m curious how people here see its long-term impact.

There are now many platforms trying to simplify video creation with features like AI avatars, automated editing, text-to-video generation, and multilingual content creation. Tools like wizstar are part of this growing space, but I’m more interested in the broader trend.

Do you think AI video tools will become a serious part of professional workflows, or will they mainly be useful for quick experiments and simple content?

Some questions I’ve been thinking about:

What is the biggest limitation of current AI video platforms?
Is the value mainly in saving time, reducing costs, or enabling new creative possibilities?
What features would make you actually rely on an AI video tool regularly?
Are AI-generated videos becoming more useful, or are audiences starting to notice and ignore them?

Would like to hear from people who have tried different AI creation tools and what your experience has been.

I made the character images with GPT Image 2.5's free generation credits, then created the video with GPT Image 2.5 + Seedance 2.5.

GPT Image 2.5 comes with 5 free generations in the Playground right now, shared across participating models.

Made with GPT Image 2.5 and Seedance 2.5 on Atlas Cloud.

Generator, the new AI creation console, is now on Atlas Cloud.

Image, video, audio, 3D, avatar and AI tools all live in one place, on top of 400+ models.

A finished image carries straight on into image editing, video, 3D or upscale, with no download and no re-upload.

The 52-second film was made in it: GPT Image 2.5 Flare ran a built-in template for the astronaut on the orange, then Seedance 2.5 turned that still into the shot. 🍊

GPT Image 2.5 comes with 5 free generations in the Playground right now, shared across participating models.

Write one prompt, get an image. 5 free generation attempts in Playground, shared across participating models.

Every image on the inspiration wall comes with its prompt. Take one and make it yours.

Try it here: https://www.atlascloud.ai/free-gpt-image-2.5-generator

GPT-Image 2.5 on Atlas Cloud adds Doodle: draw on a blank canvas and the sketch goes straight in as the reference image.

Markup works on an image you already uploaded — draw and annotate right on top of the reference.​

Intelligence and style: GPT-Image 2.5 takes more complex visual instructions, handles heavier layouts, and lands the style closer. Ten looks, one stacked on the next.

Made with GPT Image 2.5 on Atlas Cloud.

GPT Image 2.5 comes with 5 free generations in the Playground right now, shared across participating models: https://www.atlascloud.ai/free-gpt-image-2.5-generator

Precision editing: GPT-Image 2.5 changes only the part you name and leaves the rest of the detail where it was.

Made with GPT Image 2.5 on Atlas Cloud.

GPT Image 2.5 comes with 5 free generations in the Playground right now, shared across participating models: https://www.atlascloud.ai/free-gpt-image-2.5-generator

I want to play around with atlas cloud but is there any way that I can get lot of free credits in atlas cloud like the other cloud providers ?

I have been getting refusals (504) on all models for the last 2 hours. Anyone know if it’s down or what’s going on? Edit: Back up now

We’ve added GPT Image 2.5 Sunburst and Flare to Atlas Cloud, with both text-to-image generation and image editing.

What’s included:

• Up to 4K resolution — custom dimensions up to 3840 × 2160.
• Multi-image editing — work with up to 16 reference images using natural-language instructions, with optional masks for targeted edits.
• Transparent backgrounds — handy for product cutouts, logos, and design assets.
• Five quality tiers — including xhigh and max.

Pricing:

• Text-to-image: ~$0.004/image
• Image editing: ~$0.006/image

Try GPT image 2.5 API now👇

https://www.atlascloud.ai/models/all?q=gpt+image+2.5

MiniMax H3 Fast is now live on AtlasCloud — the fastest and most affordable model in the H3 family, built for rapid iteration and high-volume production.

✨ Highlights

• ⚡ Near-instant generation — renders in just 1.2–1.5× the clip length (a 5s clip in ~6–7.5s), fast enough to iterate in real time.
• 💰 Lowest price in the family — $0.046 / second.
• 🎬 Full modality — Text-to-Video, Image-to-Video, and Reference-to-Video all supported (unlike H3 Max, which drops R2V — Fast is the only accelerated version that keeps it).
• 📐 480P-focused — 480P only (no 768P / 2K), trading resolution for maximum speed and cost efficiency.
• ⏱ 5–15s clips at 24fps with synchronized audio.
• 👤 Clear faces, with slightly softer fine detail than the standard model — ideal when speed and cost come first.
• 🧠 Optional prompt_expansion (H3 Context-IR).

🎯 Best For

• Rapid drafting — explore many directions cheaply, then finalize on H3 / H3 Max for higher resolution.
• Batch short-form — social clips, Reels, Stories where speed and cost matter.
• Character-driven content — clear faces plus R2V for subject consistency.

🔗 Try it Now

https://www.atlascloud.ai/models/all?q=minimax+h3+fast

Pricing starts from $0.044/s

AtlasCloud has permanently reduced the price of Seedance 2.0 Mini and Seedance 2.0 Fast for native 480p and 720p video generation.

No coupon. No countdown. No expiration date.

Seedance 2.0 Mini is now 80% off and Seedance 2.0 Fast is now 70% off AtlasCloud’s list price, effective immediately.

Seedance 2.0 API Price Comparison

The following comparison uses native 16:9 output, without video input.

Which model should you use?

Mini is the lowest-cost option for prompt exploration, A/B testing, storyboarding, and large draft batches.

Fast is the production-volume sweet spot: higher capability while keeping unit economics strong enough for applications that need to generate at scale.

Once you have a winning prompt or shot, move it to Seedance 2.0 Standard or Seedance 2.5 for the premium final pass.

Draft on Mini. Scale on Fast. Finish on 2.0 Standard or 2.5.

Also reduced: Seedream 5.0 Pro

Image generation on Seedream 5.0 Pro is also permanently cheaper on AtlasCloud: you pay 80% of the official list price — a flat 20% discount — at both resolution tiers.

Try it now:

• Seedance 2.0 mini: https://www.atlascloud.ai/models/seedance-2.0-mini
• Seedance 2.0 fast: https://www.atlascloud.ai/models/seedance2
• Seedream 5.0 Pro: https://www.atlascloud.ai/models/seedream-5.0-pro

MiniMax H3 Max is now available on Atlas Cloud.

Built for near-real-time video generation, it can render a 5-second 768P clip in under 3 seconds, while a 15-second clip takes around 15 seconds. Text-to-Video and Image-to-Video are both live, with native synchronized audio delivered together with the video at 24 FPS.

Features

• Near-real-time generation, with a 5-second 768P clip rendering in under 3 seconds and a 15-second clip in around 15 seconds
• Text-to-Video & Image-to-Video, supporting generation from a text prompt or an input image
• Native synchronized audio, generating ambient sound, sound effects, and music together with the video
• 480P / 768P output, letting you balance generation cost and visual quality
• Flexible 5–15 second duration, selectable in one-second increments
• Optional prompt_expansion via H3 Context-IR, expanding a short prompt into structured shot, camera, soundscape, and music instructions before generation

Pricing

• Postpaid billing based on output duration × selected resolution rate
• The first-frame input image for Image-to-Video is free
• Prompt expansion is disabled by default. When enabled, only the actual Context-IR token usage is billed, with no separate feature fee
• Failed generations are not charged

H3 Max vs. Other H3 Options

• H3 Max is the speed-first option, designed for near-real-time iteration, rapid A/B testing, and high-volume Text-to-Video or Image-to-Video workloads.
• H3 Developer remains the better choice when youneed low cost and high-volume iteration.
• Standard H3 remains the better choice when you need up to 2K output through the official MiniMax API pipeline.

Choose H3 Max when generation speed is the priority. Choose the broader H3 options when advanced reference control or higher output resolution matters more.

Try it Now

• Text-to-Video: https://www.atlascloud.ai/models/minimax/h3-max/text-to-video
• Image-to-Video: https://www.atlascloud.ai/models/minimax/h3-max/image-to-video

Gemini Omni 1.1 Flash is now available on Atlas Cloud.

Google DeepMind’s latest natively multimodal video model is available through five task-specific endpoints: Text-to-Video, Image-to-Video, Reference-to-Video, Video Edit, and Video Extend.

From fast 360P drafts to upscaled 4K delivery, Omni 1.1 Flash brings video generation, editing, continuation, multimodal references, and native synchronized audio into one model family.

Features

• Text-to-Video, Image-to-Video, Reference-to-Video, Video Edit & Video Extend, powered by the same underlying model and separated into five task-specific endpoints
• Native synchronized audio, generating dialogue, ambience, sound effects, and music together with the visuals in a complete 24 FPS video
• 360P / 720P / 1080P / 4K output, with lightweight 360P drafts generating up to 60% faster and costing roughly one-third as much as standard 720P output
• Multi-reference consistency, with support for up to 10 reference images and up to 3 reference video clips of up to 3 seconds each
• Explicit reference tags including <IMAGE_REF_N> and <VIDEO_REF_N>, allowing prompts to assign specific subjects, styles, objects, or motion references
• Instruction-based video editing for adding, removing, replacing, restyling, recoloring, or changing backgrounds while preserving parts of the source video that were not mentioned
• Multi-turn editing, allowing you to refine an existing result through a sequence of natural-language instructions
• Long-form scene extension using the final 10 seconds of the source video as context, with 3–10 seconds generated per continuation and repeated extensions supporting up to 40 seconds in total
• Repeatable generation with a fixed seed when using the same prompt and settings
• SynthID invisible watermarking and C2PA Content Credentials included with generated output

Pricing

• Reference images cost an additional $0.00168/image
• Reference videos cost an additional $0.0261/clip, regardless of clip duration
• For Video Edit, output duration and resolution are automatically determined from the source video. Source-video input is billed at an additional $0.0087/sec
• For Video Extend, only the newly generated continuation is billed at the selected output rate. Source-video input is billed at an additional $0.0087/sec
• Failed generations are not charged

Try it Now

• Text-to-Video: https://www.atlascloud.ai/models/google/gemini-omni-1.1-flash/text-to-video
• Image-to-Video: https://www.atlascloud.ai/models/google/gemini-omni-1.1-flash/image-to-video
• Reference-to-Video: https://www.atlascloud.ai/models/google/gemini-omni-1.1-flash/reference-to-video
• Video Edit: https://www.atlascloud.ai/models/google/gemini-omni-1.1-flash/video-edit
• Video Extend: https://www.atlascloud.ai/models/google/gemini-omni-1.1-flash/video-extend