Abacus AI Studio - FAQs

Abacus AI Studio is your all-in-one AI creative workspace. You can generate, edit, and enhance images, videos, and speech in one chat-style workspace that brings together 50+ models from Google DeepMind, OpenAI, xAI, ByteDance, Kling, Black Forest Labs, ElevenLabs, and more, so you don’t have to jump between tools to create pro-level content.

On top of the raw models, Studio adds purpose-built tools: reusable AI Avatars, social-ready Shorts, narrated Explainer videos, and a timeline Video Editor for finishing everything you make.

Auto Mode is your easiest, fastest path. Just type what you want, and Abacus AI Studio picks the best model, resolution, and settings for the job. Static scene? It routes to image generation. Motion-heavy prompt? It routes to video generation. Ask for a voiceover, a soundtrack, or sound effects and it can generate those too.

In Auto Mode your only decision is the quality level: Standard balances quality, speed, and cost, while Premium uses the best models and maximum settings (and more credits). Switch to Image, Video, or Speech mode when you want to pick a specific model and dial in its settings yourself.

The mode selector in the prompt bar gives you six ways to create:

• Auto — Studio decides everything based on your prompt.
• Image — Pick an image model and control resolution, aspect ratio, number of outputs, and more.
• Video — Pick a video model and control duration, resolution, aspect ratio, audio, and reference inputs.
• Speech — Text-to-speech, speech-to-text, and speech-to-speech.
• Avatar — Create a social Short starring one of your AI Avatars (Pro).
• Explainer — Turn a topic into a narrated, animated explainer video (Pro).

The Explore area is a curated showcase put together by the Studio team. It’s there for inspiration, not a public feed of other users’ work, and nothing you create is ever published to it.

• Featured — A highlight reel across images, videos, Shorts formats, and Explainer styles.
• Images — Standout stills from the latest image models.
• Videos — Standout clips, with sub-tabs for All, Shorts (every social format), and Explainers (every visual style).
• Avatars — Choose a starter avatar or create your own, and browse the UGC and TikTok formats built around an avatar.

Click any tile to see its prompt, model, dimensions, and approximate credits used, then hit Recreate or Use Template to try it yourself. Your own work lives in My Recents just above Explore, and in the full My Generations gallery.

You get a stacked lineup of state-of-the-art models across image, video, speech, and sound:

• Image Models: GPT Image 2.5, GPT Image 2, GPT Image 1.5, Nano Banana Pro, Nano Banana 2, Nano Banana Lite, Seedream 5 Pro, Seedream 5 Lite, Meta Muse Image, Grok Imagine Image 2, Grok Imagine Image, Grok Imagine Quality, FLUX.2 [Pro], Midjourney, Hunyuan Image 3.0, Wan 2.7, Ideogram 3.0, Recraft SVG, Recraft Vectorize, Qwen Image Edit, and Magnific Upscaler.
• Video Models: Seedance 2.5, Seedance 2.0, Seedance 2.0 Mini, Veo 3.1, Veo 3.1 Lite, Kling AI O3, Kling AI v3, Kling AI v2.6, Kling v3 Motion Control, Kling v2.6 Motion Control, Gemini Omni Flash 1.1, Gemini Omni Flash, MiniMax H3, Hailuo 2, Wan 3.0, FLUX 3, Grok Imagine Video 1.5, Grok Imagine Video, Luma Labs, HeyGen and Hedra for lip sync, and Topaz Upscaler.
• Speech Models: ElevenLabs, OpenAI, MiniMax Speech 2.8 HD, Seed Speech, Seed Audio 1.0, Hume, and VibeVoice for text-to-speech; OpenAI for speech-to-text; ElevenLabs for speech-to-speech.
• Music & Sound: Lyria 3 Pro, ElevenLabs Music, MiniMax Music 2.6, and CassetteAI for music; Stable Audio 2.5, ElevenLabs Sound Effects, and MMAudio for sound effects.

The lineup is updated as new models ship. The model dropdown in each mode always shows what’s currently available.

Yes, you can. Hit the microphone button in the prompt bar, pick your language from the picker (Spanish, Portuguese, French, German, Japanese, Korean, Chinese, Hindi, Arabic, and many more), and speak. Abacus AI Studio turns your voice into a prompt, and your language choice is remembered for the rest of your session.

Open My Generations and click any result. Reuse prompt reloads the original prompt, model, and settings into the prompt bar so you can tweak and regenerate. Use as reference image (or video) attaches that result to your next prompt, which is the fastest way to animate, upscale, or edit something you’ve already made. Every result also has a Copy button on its prompt and a heart to save it to your Favorites.

Abacus AI Studio accepts images, video, and audio:

• Images — PNG, JPG/JPEG, WebP, GIF, HEIC, and more, up to 50 MB each.
• Video — MP4, MOV, WebM, MKV, AVI, and more, up to 200 MB and 60 minutes each. MP4 is the recommended format for Motion Control and reference videos.
• Audio — MP3, WAV, M4A, AAC, FLAC, OGG, and more, up to 200 MB each.

You can attach up to 25 files to a single prompt. Explainer mode also accepts documents so it can build a script from your own source material.

You’ve got a powerhouse image stack: GPT Image 2.5, GPT Image 2, GPT Image 1.5, Nano Banana Pro, Nano Banana 2, Nano Banana Lite, Seedream 5 Pro, Seedream 5 Lite, Meta Muse Image, Grok Imagine Image 2, Grok Imagine Image, Grok Imagine Quality, FLUX.2 [Pro], Midjourney, Hunyuan Image 3.0, Wan 2.7, Ideogram 3.0, and Recraft SVG. Each model shines in different areas like photorealism, artistic style, text rendering, reference fidelity, or vector output.

A quick guide to the lineup:

• GPT Image 2.5 — Precise instruction-following and legible text, with six quality levels and sizes up to 3840x2160.
• Nano Banana Pro / Nano Banana 2 — Google’s models with up to 4K output and optional web search grounding for factually accurate imagery.
• Seedream 5 Pro / Seedream 5 Lite — Fast, high-fidelity generation up to 3K with multi-image editing.
• FLUX.2 [Pro] and Meta Muse Image — Strong with multiple reference images for consistent subjects and styles.
• Midjourney — Signature artistic style and rich textures.
• Grok Imagine Image 2 and Ideogram 3.0 — Fast, stylish generation with solid text rendering.
• Recraft SVG — True vector output for logos, icons, and illustrations.

Not sure? Leave the model on Auto and Abacus AI Studio picks for you.

It depends on the model you choose. Most models offer 1K, 2K, and 4K output and aspect ratios from 1:1 through 21:9, including 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 4:5, and 5:4. GPT Image models let you pick explicit sizes up to 3840x2160, and Nano Banana Lite adds ultra-wide banner ratios like 8:1 and 1:8. Higher resolutions consume more credits. You’ll see the exact options in the settings pills as soon as a model is selected.

In Image and Video mode, open the “…” More settings pane to find Rewrite Prompt. When it’s on (the default for image models), AI expands your input into a richer, model-friendly prompt with better detail on lighting, composition, style, and scene clarity. Turn it off when you want the model to follow your wording literally, for example when you need exact text or a specific layout.

Absolutely. Most image models have a Number of Images setting (1 to 4) so you get several takes in a single run. You can also re-run the same prompt for fresh variations, since generation includes natural randomness. Some models expose a seed in More settings if you want reproducible results.

Recraft SVG is purpose-built for vector creation. Instead of pixel-based images, it outputs true SVG files that scale infinitely without losing quality, which makes it ideal for logos, icons, illustrations, and reusable design assets. You get more than 20 vector styles to choose from, including Engraving, Line Art, Linocut, Bold Stroke, Cutout, Editorial, Infographical, Mosaic, and Vector Photo. Recraft Vectorize goes the other way: give it any raster image, generated or uploaded, and it converts it into an SVG.

Yes. Attach reference images and describe what you want. Most models accept several references (Nano Banana up to 14, GPT Image 2.5 up to 16, Seedream 5 up to 10, FLUX.2 [Pro] up to 9) and preserve the subject’s identity across generations. In Video mode, Kling AI v3 and Kling AI O3 add Elements: named characters or objects built from 1 to 3 photos that you reference in your prompt as @Element1, @Element2, and so on. For a persistent face and voice across social videos, create an AI Avatar.

Upload an image (or pick one from My Generations), describe your edits, and let the model handle the heavy lifting. GPT Image 2.5, Nano Banana, Seedream 5, FLUX.2 [Pro], and Qwen Image Edit all support editing. You can change backgrounds, tweak colors, add or remove objects, or shift the visual style while preserving untouched regions. Most models also take multiple input images, so you can combine a product shot, a background, and a style reference into a single result.

Yes. Hover over any generated image and choose Edit Part of the Image. Paint a mask over the region you want to change, describe the replacement, and only that area is regenerated while the rest stays pixel-perfect. You can also ask Studio to expand the canvas beyond the original frame, restyle the whole image, or convert it to a vector.

Magnific Upscaler includes nine Optimized For presets so you can control how details are enhanced:

• Standard — Balanced enhancement for general use.
• Soft Portraits — Gentle skin and facial detail.
• Hard Portraits — Sharper, more defined facial features.
• Art & Illustration — Preserves stylization and brush/line character.
• Videogame Assets — Optimized for game assets and CG scenes.
• Nature & Landscapes — Enhances terrain, foliage, and environmental texture.
• Films & Photography — Adds cinematic crispness and photographic sharpness.
• 3D Renders — Sharpens geometry, lighting, and material detail.
• Science Fiction & Horror — Pushes dramatic, atmospheric detail.

You can also choose an engine (Automatic, Magnific Illusio, Magnific Sharpy, or Magnific Sparkle) to change the character of the enhancement.

You can upscale at 2x, 4x, 8x, and up to 16x. Output is capped at roughly 25 megapixels, so a 1080p source tops out around 4x while smaller sources can go higher. Final resolution depends on your input size and the factor you pick.

Yes. You can upscale any image in Abacus AI Studio, whether it was generated inside the platform or uploaded from elsewhere.

Use this winning combo: pick the preset that matches your content, describe the image in the prompt so the upscaler knows what it’s reconstructing, and start with a lower factor if your source is low-res. For fine control, open More settings and adjust the Creativity, HDR, Resemblance, and Fractality sliders. Raising Resemblance keeps the result faithful to the original, while lowering Creativity avoids invented detail.

You get a top-tier video roster: Seedance 2.5, Seedance 2.0, Seedance 2.0 Mini, Veo 3.1, Veo 3.1 Lite, Kling AI O3, Kling AI v3, Kling AI v2.6, Gemini Omni Flash 1.1, Gemini Omni Flash, MiniMax H3, Hailuo 2, Wan 3.0, FLUX 3, Grok Imagine Video 1.5, Grok Imagine Video, and Luma Labs, plus Kling v3 and v2.6 Motion Control for motion transfer, HeyGen and Hedra for lip sync, and Topaz Upscaler for enhancement. Different models excel at different things, including cinematic quality, native audio, long single takes, reference fidelity, and controlled motion.

A quick guide to the lineup:

• Seedance 2.5 — Single continuous takes up to 30 seconds, with up to 30 reference images plus reference videos and audio. Great for consistent characters and products.
• Seedance 2.0 — Up to 4K output with fast and standard modes.
• Veo 3.1 — Cinematic quality with native audio, up to 4K. It’s also the most credit-intensive model. Veo 3.1 Lite is the lighter, cheaper option.
• Kling AI O3 / Kling AI v3 — Elements for character consistency, shot-structure control, and a 4K mode on O3.
• Gemini Omni Flash 1.1 — Fast and low cost with native audio, up to 4K, and the best pick for quick video-to-video edits.
• MiniMax H3 — Up to 4K with native audio and multiple reference inputs.
• Wan 3.0 — Clips from 2 to 30 seconds that can pull context straight from a webpage or document URL.
• FLUX 3 — 5 to 20 second clips and native extension of an existing video.
• Grok Imagine Video 1.5, Luma Labs (with seamless looping), and Hailuo 2 — Fast, stylish options for shorter clips.

Not sure which one to use? Auto Mode has your back and picks for you.

Duration options are model-dependent. Most models generate 1 to 15 second clips, with Seedance 2.5 and Wan 3.0 producing single takes up to 30 seconds and FLUX 3 up to 20 seconds. Veo 3.1 offers 4, 6, or 8 seconds, and Auto Mode lets you pick anywhere from 3 to 15 seconds. Need something longer? Just ask. Abacus AI Studio plans a multi-scene video, shows you the plan and a credit estimate, and generates it once you approve.

Resolution support varies by model. Most offer 480p, 720p, and 1080p, and select models support 4K: Veo 3.1, Seedance 2.0, MiniMax H3, Gemini Omni Flash 1.1, and Kling AI O3 in 4K mode. Higher resolutions consume more credits, and video edits are capped at 1080p.

Supported options include:

• 16:9 — Standard widescreen, supported everywhere.
• 9:16 — Vertical format for Reels, TikTok, and Shorts, supported everywhere.
• 1:1 — Square format on most models.
• 4:3, 3:4, and 21:9 — Available on Seedance, MiniMax H3, Wan 3.0, FLUX 3, and Luma Labs.

Motion Control and lip-sync outputs follow the shape of your input image or video.

Yes, every workflow is supported:

• Text-to-Video — Type your scene and generate from scratch.
• Image-to-Video — Upload a still image as the start frame and animate it.
• Start + End Frame — Give both frames and let the model fill in the motion between them.
• Reference-driven — Attach multiple images, video clips, or audio and describe the scene. Seedance 2.5, Wan 3.0, MiniMax H3, Kling AI O3, and Gemini Omni Flash 1.1 all support this.
• Video-to-Video — Upload a clip and describe the transformation, from restyling to swapping objects.

Many models generate audio natively. Veo 3.1, Gemini Omni Flash, Grok Imagine Video, and MiniMax H3 always produce sound, while Seedance, Kling AI v3, Kling AI O3, Wan 3.0, and FLUX 3 have a Generate Audio toggle (the speaker pill in the prompt bar). Luma Labs and Hailuo 2 generate silent clips. You can always add a voiceover, music, or sound effects afterwards in Speech mode or the Video Editor.

Yes. Ask for “make it longer” or “continue the scene” and Abacus AI Studio extends it, using the last frame as the start of the next clip or, on FLUX 3 and Seedance 2.5, extending the original clip natively. For multi-scene stories, describe the whole thing and Studio plans and renders each scene, then stitches them together. You can fine-tune the cut in the Video Editor.

Motion Control (powered by Kling v3 Motion Control and Kling v2.6 Motion Control) transfers the movement from a reference video onto a character of your choice. Upload a character image and a reference video showing the motion you want, and the model makes your character perform that movement, camera behavior included. You can keep the reference video’s original sound or drop it.

• Reference Video — Gives frame-by-frame motion guidance: what moves, how, and when.
• Reference Image — Gives a visual anchor for appearance: who or what is in the scene and how it looks.
• Character Image (Motion Control) — The subject that will perform the reference video’s motion.

Motion Control reference videos can be 3 to 30 seconds long (up to 10 seconds when you choose to match the character image’s orientation). For reference-driven generation, Seedance 2.5 accepts up to 30 seconds of combined reference footage, Wan 3.0 up to 15 seconds, and Gemini Omni Flash 1.1 up to 3 seconds per clip. Videos you upload for editing can be up to 60 minutes long.

Pick HeyGen (Lip Sync) or Hedra (Lip Sync) from the Video model dropdown. Upload a portrait image (Hedra can also generate a character from a description), then either type a script and choose a voice, or upload your own audio. Abacus AI Studio generates the speech and syncs lips, jaw movement, and facial expressions to match naturally. HeyGen outputs 16:9 or 9:16 at 720p or 1080p; Hedra supports 1:1, 16:9, and 9:16.

HeyGen comes with 12 built-in voices grouped by style: Professional, Natural, Newscaster, Calm, Friendly, Deep, Bright, and Authoritative. You can preview each one before generating. Hedra takes free-text voice instructions instead. For anything more specific, generate the speech first in Speech mode (ElevenLabs, OpenAI, MiniMax, and more) and upload the audio.

HeyGen has an Expressiveness setting (Low, Medium, High), and Hedra accepts voice instructions describing how the line should be delivered. For full control, generate the audio in Speech mode, where OpenAI and Seed Speech take voice instructions, MiniMax offers emotion presets, and Hume lets you describe the voice you want, then use that audio for the lip sync.

Topaz Upscaler lets you choose any factor from 1x to 4x, and picks from 19 enhancement models (Proteus by default, plus the Artemis, Gaia, Nyx, and Starlight families) tuned for different kinds of footage. You also get sliders for noise reduction, detail recovery, compression-artifact removal, halo reduction, and film grain.

Yes. Turn on Frame Interpolation and set a target FPS between 16 and 60. Raising the frame rate above the source doubles the cost of the upscale.

The Video Editor is a timeline editor built into Abacus AI Studio for finishing what you generate. Open it from the Video Editor link under the prompt bar, or from the Editor button inside any Studio conversation. Import your generations and uploads, then cut, split, trim, reorder, speed up, crop, and mix audio; add transitions and text overlays; and generate auto captions from the audio in multiple languages (or upload your own .srt). Landscape, Portrait, and Square presets cover YouTube, Reels, TikTok, and Instagram. Exports render in the cloud and land in My Generations, with version history so you can always go back. You can also ask Studio to make edits for you (“cut the first two seconds”, “add captions”) and it updates the project. The editor is desktop only, since it needs a wide screen.

Yes. Upload a clip and ask. Trimming, cutting, speed changes, cropping, joining clips, extracting a frame, or pulling out the audio track are all handled as direct edits rather than AI generation, so they’re fast and inexpensive.

Speech mode covers three workflows:

• Text to Speech — Turn a script into natural voice audio with your choice of model and voice.
• Speech to Text — Transcribe an audio file (up to about 23 minutes).
• Speech to Speech — Re-voice a recording (up to 5 minutes) in a different voice, with optional background-noise removal.

Leave the sub-mode on Auto and Studio picks the right one from your prompt and attachments.

You get ElevenLabs (the full voice library, with Flash, Multilingual, and V3 sub-models across 30+ languages), OpenAI (13 voices plus free-text voice instructions), MiniMax Speech 2.8 HD (17 voices with emotion presets), Seed Speech (voice instructions), Hume (describe the voice you want), VibeVoice (English and Chinese speakers), and Seed Audio 1.0 (describe the audio you want, including tone and delivery). Speech-to-text runs on OpenAI and speech-to-speech on ElevenLabs.

Yes. Ask for it in Auto Mode. For music you get Lyria 3 Pro, ElevenLabs Music (up to 10 minutes), MiniMax Music 2.6 (with lyrics), and CassetteAI (instrumental). For sound effects you get Stable Audio 2.5, ElevenLabs Sound Effects, and MMAudio, which generates sound matched to a video. Layer the results onto your video in the Video Editor.

Shorts are finished, social-ready vertical videos built from proven formats and starring one of your AI Avatars. Pick from 24 formats across three categories:

• UGC — Selfie Review, Tutorial, GRWM, Unboxing ASMR, Before & After, Virtual Try-On, Cozy Vibes, and Penthouse Flex.
• TikTok — Street Interview, Listicle, Comparison Test, Secret Hack Reveal, This Gadget Saved Me, Camera POV, Red Carpet Arrival, Hero Transformation, POV Thrill, Breaking News, Stadium Walkout, Wild Card Commercial, and Wild Card Short.
• Commercial — TV Spot, Hyper Motion, and Survival Test.

Eight of these are Spotlight formats that star your avatar with no product at all: Red Carpet Arrival, Cozy Vibes, Hero Transformation, Penthouse Flex, POV Thrill, Breaking News, Stadium Walkout, and Wild Card Short. Don’t see the format you want? Skip the template and describe the short yourself.

Switch the mode selector to Avatar (or open Videos → Shorts in Explore) and follow the steps:

• Step 1: Pick an avatar. Choose a starter avatar or one you’ve created.
• Step 2: Optionally pick a template, and add your product with a URL or photos.
• Step 3: Describe what you want, then set the quality (Standard on Seedance 2.0 for up to 15 seconds, or Premium on Seedance 2.5 for a single take up to 30 seconds), resolution up to 1080p, and aspect ratio (9:16 by default).
• Step 4: Review the plan. Studio writes the script and shot list and shows a credit estimate. Hit Accept & generate, or Make it cheaper to re-plan at Standard quality. Any follow-up message revises the plan before you commit.

The result is rendered as one continuous take with your avatar’s face and voice, ready to post.

An AI Avatar is a reusable persona with its own face and voice. Create it once and every Short you make with it keeps the same look and sound.

To create one, open the Avatars tab and click Create an Avatar. For the look, either describe the person and let Studio generate a portrait (you can attach up to 3 reference images to steer it, then edit the result in place with instructions like “give her a denim jacket”), or upload up to 3 photos of your own. For the voice, describe how they sound (age, accent, tone, pace) and audition the result, or upload a 30 to 60 second recording. Starter avatars are available if you want to skip creation entirely.

Avatars must be original characters. Descriptions of real or identifiable people and public figures are declined.

Yes. Paste a product URL and Studio pulls the name, details, and images (app store links are recognized and use the app’s screenshots), or upload your own product photos. Studio generates clean studio views of the product and builds the short around them, so it looks right in every shot. You can add up to 3 links and 3 documents per product.

Explainers turn a topic into a fully narrated, multi-scene animated video. Studio researches the topic if needed, writes the script, records the voiceover with one of 20 curated voices, generates every scene in your chosen visual style, adds a background music track, and stitches it all together.

You choose from 12 visual styles (Pixel Art, Claymotion, Mixed Media, 3D Papercraft, 2D Illustrator, Whiteboard Doodle, Low Poly, 3D Mix, Anime, Watercolor, Comic Book, and Isometric 3D), a length from 20 seconds up to 10 minutes, and 16:9 or 9:16. Upload up to 3 documents to have the script built from your own material.

Switch the mode selector to Explainer (or open Videos → Explainers in Explore), pick a style and voice, type your topic, and send. Before anything renders, Studio shows a plan card where you can play the narration, read the script, review the scene-by-scene breakdown, preview the music, and see the credit estimate. Hit Accept & generate when you’re happy. To change something, just reply: “make it one minute”, “make it vertical”, or “rewrite scene 3 to focus on pricing”. Scenes that didn’t change are reused at no extra cost. Add captions afterwards in the Video Editor.

These are Pro features. They’re included on the Pro and Max tiers and for enterprise accounts, and unlocked by your organization’s admin. Basic subscribers can browse the Shorts formats and Explainer styles in Explore and will see an upgrade prompt when they try to create one. As with everything in Studio, you see a credit estimate on the plan card and only pay when you accept it.

The paperclip in the prompt bar is your hub for reference assets. You can:

• Upload from computer — Pull in images, videos, or audio directly from your device.
• Add from files — Pick from anything you’ve previously uploaded to Abacus.AI.
• My Generations — Browse your own Studio results in a grid or list, filter by type, search, and multi-select in the order you want them used.
• Connect Apps — Import from Google Drive, OneDrive, Box, or SharePoint without downloading first.

You can also paste images straight from your clipboard or drag and drop files onto the prompt bar.

Yes. Open the attachment menu, choose Connect Apps, and connect Google Drive, OneDrive, Box, or SharePoint. Once connected, you can browse and import image, video, and audio assets into your workflow without downloading them first.

Chaining is native to the conversational experience. Everything you make stays in context, so just keep going step by step:

• Step 1: Generate an image with any image model.
• Step 2: Say “upscale that 4x with Magnific” for more resolution and detail.
• Step 3: Say “animate it into a 10 second clip” and Studio uses it as the start frame for an image-to-video model.
• Step 4: Say “make it longer” to continue the scene, then “upscale it with Topaz at 60 fps” for the final polish.
• Step 5: Say “open it in the Video Editor and add captions” to finish and export.

You can also attach any earlier result explicitly with Use as reference from My Generations. Studio shows a credit estimate before each step, and asks you to confirm before an especially expensive one.

Everything you create is private by default, and nothing is ever published to the Explore gallery. To share a result, open it from My Generations and click Share. You can restrict it to specific people, open it to anyone in your organization, or create a public link that works without signing in. You can make it private again at any time, and delete any generation permanently from My Generations.

Click the Image / Video tile in the left navigation (or the Studio pill at the top of a chat). Your Studio conversations are listed under Studio in the left navigation, separate from your regular chats. If you ask ChatLLM or Abacus AI Agent for a video, it can offer to continue in Studio with your prompt carried over.

Abacus AI Studio is included with your ChatLLM subscription. What you can do depends on your tier:

• Basic Tier — $10/month, 20,000 credits: Unlimited image generation, video generation in up to 3 conversations per billing month, and up to 2,500 credits per conversation.
• Pro Tier — $20/month, 30,000 credits: Everything in Basic with no video-conversation or per-conversation caps, plus Shorts, Avatars, and Explainers. If you’re already on Basic, Pro is an extra $10/month.
• Max Tier — $100/month, 120,000 credits: Everything in Pro with four times the credits.

The Go tier ($7/month) does not include Studio. Studio is also available through ChatLLM Teams subscription plans.

Credit usage depends on the model, resolution, duration, quality level, number of outputs, and any reference media you attach. You’ll see an Estimated Cost in the prompt bar before you generate, and for especially expensive requests Studio shows the estimate with a Generate / Cancel choice before it starts. Explore tiles also show the approximate credits used to make each example.

Yes. Higher resolutions require more compute, so they consume more credits, and 4K options are flagged in the settings. A good habit is to draft at 720p or 1K and only go up for the final version.

Yes. Video credit usage scales with duration, roughly per second on most models, and some models price longer clips at a higher rate. Standard quality is cheaper than Premium at the same length.

Upscaling is its own generation step, so it uses additional credits. Magnific pricing scales with the factor you choose, and Topaz pricing scales with output resolution and factor, doubling if you turn on Frame Interpolation to raise the frame rate.

Yes. Every Studio response shows the credits it used, and clicking that label opens your profile, where the billing dashboard has your real-time balance and a usage history you can filter by period and break down by conversation.

Studio warns you when your balance is running low. If it hits zero, you won’t be able to start new generations until you add more credits. You can buy credits any time from the billing page, or turn on auto-refill so you never run dry mid-project.

If a generation fails because of an error on our side, the credits for that turn are refunded automatically and Studio tells you so in the chat. Failures caused by the input itself (for example a prompt a model’s safety system declines) or by hitting a plan limit aren’t refunded, but Studio will usually suggest how to adjust and try again.

Commercial rights depend on each model’s license terms. In general, content from Abacus AI Studio can be used commercially, but certain model-specific restrictions may apply.