AI Model Guides

GPT Image 2.5 Flare is DayGen's default GPT image model. It renders higher-quality images than GPT Image 2 at roughly half the latency, so it suits everyday creation, social and creator content, product imagery, and high-volume work. It keeps the full GPT Image contract: up to 16 references, masked inpainting, streaming previews, and it adds the Extra high and Max quality tiers plus transparent background output.

GPT Image 2.5 Sunburst is the premium member of the GPT Image 2.5 family. It trades generation time for tighter control across edits, which suits production-ready campaign creative, polished product imagery, and detailed multi-reference work. It shares Flare's contract: 16 references, masked inpainting, streaming previews, Extra high and Max quality tiers, and transparent background output.

Nano Banana Lite is the quickest and lowest-cost Gemini image option in DayGen. Use it for first passes, everyday drafts, quick edits, and up to 14 object or general references. Its DayGen reference budget does not admit Avatar, Character, or style identity roles.

Use MAI Image 2.5 Standard for polished portraits, readable labels and posters, product imagery, brand assets, commercial layouts, and precise prompt-directed edits through FAL.

Use MAI Image 2.5 Pro for production-grade portraits, products, typography, brand work, commercial design, and precise one-image edits when the strongest MAI finish matters.

Create text-led visuals or edit an image with written instructions and up to three ordered references. Qwen Image 3.0 is an optional model in Other models, with no change to your default.

Use Meta Muse for precise instruction following, readable text, plots and QR codes, or for composing and editing with up to ten ordered image references through FAL.

Nano Banana 2 is best for most image tasks, especially when Avatar/Characters or multiple references matter. Its DayGen budget supports up to 10 object or general references plus four Avatar/Character references, but no style-role references.

Nano Banana Pro is the premium Gemini choice for images that need more realism, stronger face detail, and a more finished look. Its DayGen budget supports up to six object or general references, five Avatar/Character references, and three style-role references.

Grok Imagine Image 2.0 is xAI's current image model in DayGen for expressive creation, ordered multi-image editing with up to three references, and polished social or campaign visuals at 1K or 2K. Choose Low for faster drafts or Medium for the default quality tier.

FLUX.2 [max] is the highest-quality FLUX.2 option in DayGen. It creates or edits images with up to eight ordered references and supports output up to four megapixels, making it the family choice for final-detail work.

FLUX.2 [pro] balances image quality and production speed for everyday FLUX.2 work. DayGen uses Black Forest Labs' current preview endpoint and supports creation or editing with up to eight ordered references and output up to four megapixels.

FLUX.2 [flex] is the most adjustable FLUX.2 option in DayGen. It supports up to eight ordered image references, output up to four megapixels, guidance from 1.5 to 10, and one to 50 inference steps.

FLUX.2 [klein] 9B is the stronger Klein option for fast FLUX.2 drafts and edits. DayGen uses Black Forest Labs' current preview endpoint and accepts up to four ordered image references.

FLUX.2 [klein] 4B is the fastest and lowest-cost public FLUX.2 API option in DayGen. Use it for rapid drafts and edits with up to four ordered image references before moving to a higher-quality family member.

Luma Uni-1 is strongest when you want cinematic images with controlled edits, angle changes, and visual continuity across references. Use it for image creation, edits, resize, outpaint, and angle work where the output should feel like a controlled continuation of the source material.

Luma Uni-1 Max is the higher-fidelity Luma option. Use it when reference consistency and image quality both matter: polished fashion shots, product visuals, cinematic portraits, and final angle or resize outputs.

Seedream 5 Pro is the official Seedream 5 professional image model in DayGen. It is built for polished prompt-to-image and reference-guided image generation, including multi-reference product, character, place, logo, and style workflows.

Seedream 5 Lite is a fast all-around image model for structured prompts, quick polished drafts, and broad creative range. It is the least restrictive image option in DayGen while still supporting references and instruction-heavy prompts.

Ideogram 4.0 is the model to try when the image includes readable words, poster text, logos, packaging text, signage, or graphic layouts. It is less about broad multi-reference editing and more about controlled text-in-image generation.

Ideogram Reframe expands or crops one source image into an exact canvas size while rebuilding the surrounding composition. Use it when the content already works but must fit a different placement.

Ideogram Replace Background keeps the foreground of one source image and generates a new background from your prompt. Use it for product, portrait, and campaign composites where the subject should survive the edit.

Ideogram Ad Resizer recomposes one finished advertisement for an exact placement size. Use it to adapt campaign creative across banners, feeds, stories, and other channel-specific canvases.

Reve 2.1 is useful when several references need to hold together in one image. Use it for reference-heavy scenes, consistent visual direction, product or character ideas, and controlled creative blends.

Reve 2.1 Layout separates composition from rendering so follow-up edits can preserve and reshape the image structure. Use it for posters, editorial layouts, controlled scenes, and iterative composition work.

Recraft Remove Background is DayGen's fixed background-removal tool. It sends the source image to Recraft's native cutout endpoint and returns a derived image with transparency while preserving the source asset lineage.

Recraft V3 Styles turns a prompt into a raster image using one of DayGen's curated Recraft looks. Use it when a ready-made illustration, photography, or graphic style is more useful than reference-image control.

Recraft V4.1 is strongest when the output should feel like a designed asset. Use it for brand graphics, icons, clean marketing visuals, stylized product shots, and layout-aware images.

Recraft V4.1 Pro is the stronger Recraft choice when the design asset needs more resolution headroom. Use it for campaign visuals, print-adjacent imagery, posters, and polished design work where crisp output matters.

Recraft Utility is the practical Recraft option for clean, controlled image ideas that do not need the full Pro resolution tier. Use it for everyday design assets, illustrations, and simple branded concepts.

Recraft Utility Pro gives the practical Utility style more resolution headroom. Use it for general design assets that need cleaner detail, larger output, or more room for later cropping.

Recraft V4.1 Vector generates editable SVG artwork from a prompt. Use it for icons, logos, illustrations, and clean graphics that need to remain scalable and easy to refine after generation.

Recraft V4.1 Pro Vector is the premium SVG generation tier for polished brand assets, logos, icons, and production graphics where the final vector needs more refinement.

Recraft V4.1 Utility Vector is the practical SVG option for everyday icons, diagrams, illustrations, and design drafts that need editable vector output.

Recraft V4.1 Utility Pro Vector keeps the general-purpose Utility behavior while adding the premium tier for cleaner production-ready SVG assets.

Recraft V3 Vector combines the V3 style library with editable SVG output. Use it for stylized illustrations and graphics that need both a ready-made look and scalable vector paths.

Recraft Vectorize converts one raster source image into editable SVG paths. Use it when the artwork already exists and the goal is a scalable vector derivative rather than a new prompt-generated design.

Krea 2 Medium is a strong creative exploration model. Use it when mood, palette, texture, style, and aesthetic interpretation matter, especially with style references or a moodboard-like brief.

Krea 2 Large is the quality-focused Krea option. Use it when the creative direction is set and you want more detail, polish, and aesthetic depth from the Krea family.

Krea 2 Medium Turbo is for fast Krea iteration. Use it to search through aesthetics, color stories, style references, and composition ideas before spending more time on a higher-quality pass.

Use this Ideogram path when the active DayGen edit flow is specifically a masked inpaint operation. For normal text-to-image and typography-first generation, use Ideogram 4.0 instead.

Magnific Creative is for turning a finished image into a more detailed, higher-resolution version with some creative enhancement. Use it when you want richer texture, detail, and upscale character, not a perfectly conservative resize.

Magnific Precision is the conservative upscale choice. Use it when the source image already works and the goal is cleaner resolution, sharper detail, and fewer creative surprises.

FLUX 3 creates five-to-20-second HD or Full HD video from text, a Start Frame, Start and End Frames, or a source video for continuation, with eight aspect ratios and optional native audio at 24 fps.

Seedance 2.5 is DayGen's reference-rich cinematic video model, with 480p/720p/1080p output, native audio, and up to 30-second output in Create.

Seedance 2.0 is the most flexible video generator in DayGen when you have more than a text prompt. Use it for first/last frames, reference images, short reference videos, audio cues, and structured multimodal direction.

Seedance 2.0 Fast keeps the Seedance multimodal workflow but prioritizes speed and lower-resolution draft work. Use it when you need to test references, motion, or audio direction before committing to the standard model.

Seedance 2.0 Mini keeps the Seedance frames, image references, video references, and audio reference workflow while prioritizing lower-cost 480p/720p generation.

MiniMax H3 creates fixed-2K video from a required prompt plus either first/last frames or up to 12 mixed references. Its reference mode supports up to nine images, three videos, and three audio clips within that shared limit.

H3 Max by fal is a separate fast MiniMax H3 variant for text-to-video or start/end-frame video with native synchronized audio. It supports 480p or 768p output and clips from 5 to 15 seconds.

H3 Max Turbo by fal is the fastest, lower-cost H3 Max variant for text-to-video or start/end-frame video with native synchronized audio. It supports 480p or 768p output and clips from 5 to 15 seconds.

H3 Max Extend Video by fal continues a source clip by 5 to 15 seconds while keeping its look. The source can be 1.625 to 60 seconds and up to 50 MB. fal returns the original clip followed by the new footage at 480p, 768p, 1080p, or 2K.

LTX 2.5 Fast creates landscape or portrait video from text, an opening frame, opening and ending frames, or a short audio reference with an optional opening image.

Kling 3 is best when the prompt itself is the main director. Use it for cinematic shots, camera moves, and multi-shot ideas where text direction matters more than a large reference set.

Kling 3 Omni is the Kling model for reference-rich work. Use it when images or a short video reference should guide the output, especially for controlled subject, style, or motion continuity.

Gemini Omni 1.1 Flash is Google’s stable multimodal video model in DayGen. It creates 3–10 second native-audio clips from text, first/last frames, up to six visual inputs, or up to three short video guides, with 360p through upscaled 4K output and stateful edits or extensions.

Grok Imagine Video remains available for short source-video edits and extensions that Grok Imagine 1.5 does not support. New text, image, and reference generation uses Grok Imagine 1.5.

Grok Imagine 1.5 is DayGen's Grok model for new video generation. Start from text, animate one still, or guide a clip with two to seven visual references while keeping native audio.

P-Video is a practical speed model for drafts. Use it to test motion, timing, or audio-conditioned ideas quickly before spending more credits or time on a premium video model.

Happy Horse 1.1 is useful when video and native audio should feel designed together. Use it for short cinematic or social clips, reference-driven character scenes, product demos, and edits with ordered image references.

Veo 3.1 Standard is the premium Google video choice for cinematic text-to-video and image-to-video. Use it when you want strong realism, camera language, and native audio support where DayGen exposes it.

Veo 3.1 Fast keeps the Veo family workflow but trades some quality headroom for speed. Use it to explore cinematic prompts, image-to-video setups, or Veo-native audio direction before choosing a final pass.

Veo 3.1 Lite keeps the Veo create workflow for text-to-video and image-to-video while focusing on lower-cost 720p or 1080p generations. Use it for early concepts, quick alternates, or budget-conscious Veo tests.

Runway Gen-4.5 is a strong choice for controlled short clips from text or images. Use it when you want Runway's cinematic look, clear prompt-to-shot behavior, and start/end frame guidance for concise video generation.

Runway Model Router lets an opt-in Runway policy choose an eligible generation model for a prompt or Start Frame request. DayGen dry-runs the request before enqueueing and bills the model the router selects.

Runway Aleph 2.0 is for transforming existing footage. Use it when you already have a source clip and want to restyle, alter, or reinterpret it with a text prompt and optional image guidance.

Runway Ruby HDR converts one SDR source clip into a true HDR deliverable without requiring a prompt. Use it when the edit is a color-space and delivery conversion rather than a creative video transformation.

Luma Ray 3.2 is a premium option for cinematic, polished short video and creative restyling. Use it when you want Luma's motion quality, start/end frame control, or source-video restyle behavior.

Create videos up to 30 seconds from text, start and end frames, or mixed image, video, and audio references—with native sound.

Use the faster Wan 3.0 tier for time-sensitive video creation, with the same text, Frames, References, and native-audio controls as standard Wan 3.0.

Wan 2.7 is an image-to-video specialist. Use it when a still image should come alive, when first and last frames define the transition, or when an audio reference should guide motion timing.

Kling Motion Control is for directing movement rather than just prompting a clip. Use it when the camera path, motion transfer, or action control is the important part of the output.

Kling Avatar is for turning an image and voice or audio direction into an animated talking-head style video. Use it when the output is a person or avatar speaking, not a general cinematic scene.

Topaz Video is a finishing tool, not a generator. Use it after you already have a clip and need upscaling, denoising, face/detail recovery, interpolation, or cleanup before final delivery.

Enhance a private video with Runway Magnific while keeping its DayGen source and ownership lineage intact.

ElevenLabs v3 is DayGen's main voice model when performance matters: narration, character reads, ads, audiobooks, expressive dialogue, and cloned or library voices.

Gemini 3.1 Flash TTS is useful when the prompt should direct how speech is performed. Use it for narrated lines where style, mood, pacing, or a named Gemini voice matters more than choosing from a large cloned voice library.

Grok TTS is useful for quick expressive reads and voice experiments. Use it when you want a different voice flavor from ElevenLabs or Gemini and do not need the ElevenLabs clone/library workflow.

Lyria 3.5 is DayGen's first-choice Google music model for complete songs. It follows prompts, image inspiration, lyrics, structure, and target-duration guidance, with MP3 or WAV output.

Lyria 3 Preview is for quick musical ideas: hooks, loops, mood tests, short beds, and prompt exploration. Use it when you need a fast 30-second musical direction before generating a full track.

Lyria 3 Pro is the earlier full-track model retained for existing projects. It supports structured songs, image inspiration, and MP3 or WAV output.

MiniMax Music 3.0 is for complete songs up to five minutes. Use your own section-tagged lyrics, let the model write lyrics from the prompt, or choose instrumental output.

ElevenLabs Music is the flexible-duration music option. Use it when you need a track, loop, ambience, or bed with a specific length, clear style language, and simple prompt-driven music direction.

Seed Audio is ByteDance's voice and audio-scene model. Use it for dialogue, narration, character voice scenes, ambience, and sound design from text, audio references, or an image reference.

Claude Opus 5.5 is the Claude option for complex agentic planning, long briefs, and multi-step creative work. Use it when you want deep reasoning at a lower planning cost than Fable 5.

Claude Fable 5.1 is a high-capability Claude option for ambitious, long-horizon, multi-step planning. Use it when a brief needs careful structure across a longer creative arc.

Claude Fable 5 is the Claude option for ambitious, story-heavy, or multi-step planning. Use it when the brief needs careful structure, tone, sequencing, or narrative shape.

Claude Sonnet 5 balances strong agentic planning with speed and cost. Use it for dependable multi-step creative work, structured briefs, and everyday campaign planning.

The creative plan has many dependent steps.

GPT-6 Sol balances strong agentic planning with speed and cost for complex creative work.

GPT-6 Luna is the fast, low-cost OpenAI planner for focused briefs and high-volume iteration.

Gemini 3.8 Flash is the newest Google planner in DayGen. Use it for the strongest Gemini multimodal reasoning, long multi-step creative plans, and briefs that mix visual and audio references.

Gemini 3.7 Flash is the primary Google planner for visual and audio tasks. Use it for capable prompt structuring, multimodal reasoning, and reliable multi-step creative planning.

Gemini 3.5 Flash-Lite is the fast Google option for concise briefs and lightweight planning. Use it when speed and efficiency matter more than deep, long-form reasoning.

Grok 4.6 is xAI's frontier Agent model for long-running, multi-step planning with text and image context. Use it for ambitious creative work that benefits from sustained reasoning and a direct planning voice.

DeepSeek V4.1 Flash is a fast, economical Agent planner for coding, reasoning, and long-running tool work. Use it when you want strong agentic planning without the cost of a specialist model.

GLM 5.3 Flash is an ultra-efficient multimodal Agent planner with long context. Use it for high-volume planning, agent loops, and coding-minded drafts.

Muse Spark 1.3 is Meta's Agent planner for long-running, multi-step creative work. Use it when a brief needs sustained agentic planning rather than a short one-pass answer.

GLM 5.3 is Z.ai's Agent planner for coding-heavy and structured agent work. Use it when the brief needs precise tool sequencing or implementation-minded planning.

Kimi K3 is a specialist Agent planner for difficult long-horizon, repository, and multimodal work. Use it when model quality justifies a higher output cost.

Qwen 3.8 Max is Alibaba's Agent planner for mixed text and visual briefs. Use it when the plan needs multimodal reasoning across media types.