# BlendCanvas — AI Image Generation & Infinite Canvas Reference > BlendCanvas is the AI image and video workspace at blendcanvas.com. Connect product references, compare image variations, and keep video takes on the same canvas. Canonical website: https://blendcanvas.com/ Simplified Chinese: https://blendcanvas.com/zh/ Traditional Chinese: https://blendcanvas.com/zh-tw/ French: https://blendcanvas.com/fr/ Spanish: https://blendcanvas.com/es/ Workspace: https://app.blendcanvas.com Official subscription and credit pack prices: https://blendcanvas.com/pricing/ This document summarizes the public BlendCanvas website (https://blendcanvas.com/, operated by Quguo Limited; distinct from standalone background-removal apps named Blend). Model availability, prices, welcome credits, and purchase availability can change. Read the live public catalog at https://app.blendcanvas.com/api/catalog and confirm the rate shown in the workspace before generating. This static reference deliberately does not quote numerical rates or promotional allowances. ## Product facts ### What can I create? Generate and edit AI images on an infinite canvas, turn frames into video, create music, and keep prompts, references, and branches connected. ### How do I get started? Sign in, drop reference images or a brief onto the canvas, choose a cloud image or video model, and generate variations side by side. ### How is generation billed? Generations use shared account credits based on the selected cloud model and resolution tier. No personal API keys or local GPU setup are required. ### Where is my work saved? Accounts and canvas node graphs are saved in MySQL, media files are stored in cloud object storage (S3/R2), and browser storage is an account-isolated cache. ## Canvas-native AI image & workflow capabilities - **Node-connected references**: Wire product photos, lighting references, character sheets, and text briefs directly into image or video generation nodes so context stays visible. - **Multi-model image comparison**: Branch one prompt or reference across multiple cloud image models and resolution tiers to compare composition, typography, and material fidelity side by side. - **Mask inpainting & smart upscale**: Refine specific regions with canvas mask editing, upscale selected frames to higher resolution, or crop and split multi-panel generations without leaving the board. - **3D camera & angle control**: Adjust subject framing and camera perspective with the built-in 3D director and angle tools before generating consistent multi-angle shots. - **Cloud Canvas Agent**: Ask the cloud-hosted assistant to turn a short brief into a structured character turnaround sheet, storyboard sequence, or multi-branch moodboard. - **Remote MCP integration**: Connect Claude Code, Codex, Cursor, or other MCP clients over Streamable HTTP with a personal token to inspect and draft canvas nodes. ## Features ### AI image generation canvas (https://blendcanvas.com/features/ai-image-workspace/) Generate, compare, and edit AI images on an infinite canvas. Connect product photos or style references, branch across multiple cloud image models, and refine details with mask inpainting, camera angle control, and upscaling. 1. Place a creative brief and reference images (product, lighting, palette, or character) onto the canvas. 2. Connect inputs to one or more cloud image models and select your aspect ratio and resolution tier. 3. Compare generated variations side by side, then refine winning frames with mask inpainting, angle adjustment, split, or upscaling. - **Multi-reference image wiring**: Keep product silhouettes, color palettes, and composition references visibly linked to every generated variation. - **Side-by-side multi-model shootout**: Test the same prompt across different cloud image models on one board instead of switching tabs or re-uploading assets. - **Mask edit, upscale & split**: Paint masks for targeted inpainting, split multi-grid sheets into standalone image nodes, crop framing, and upscale final picks. - **3D director & camera angles**: Preview camera angles on a 3D mannequin stage to guide consistent perspective across hero, detail, and turnaround shots. ### Infinite canvas for AI creative workflows (https://blendcanvas.com/features/infinite-canvas/) Organize briefs, reference boards, AI images, video clips, and audio on a cloud-synced infinite canvas—with an AI Canvas Agent and remote MCP support, without local GPU or ComfyUI setup. 1. Sign in and create a cloud-synced canvas for your campaign, storyboard, or moodboard. 2. Connect text, image, video, and audio nodes manually—or ask the Canvas Agent to lay out a complete workflow graph. 3. Branch experiments non-destructively, duplicate proven boards for new products, and access canvases via remote MCP. - **Non-destructive visual branching**: Explore lighting, copy, and framing variations in parallel branches while keeping original references untouched. - **Canvas Agent graph builder**: Let the cloud assistant draft scripts, character sheets, storyboards, and connected node layouts from one natural-language brief. - **Zero local GPU or API key setup**: Run multi-step visual workflows in the browser using managed cloud models and one shared credit balance. - **Cloud sync & hosted MCP**: Projects sync to MySQL and cloud object storage, with authenticated remote MCP tools for external AI coding and writing clients. ### AI image-to-video & storyboard workspace (https://blendcanvas.com/features/ai-video-workspace/) Turn canvas-generated frames or uploaded references into AI video clips. Keep source keyframes, motion prompts, first/last frame controls, and generated takes side by side on the BlendCanvas canvas. 1. Generate or upload a clean hero keyframe on your canvas. 2. Wire the image node into a compatible cloud video model, set aspect ratio and duration, and describe camera and subject motion. 3. Compare multiple motion takes next to your original still frame and audio tracks on the same board. - **Image-to-video continuity**: Pass approved canvas images directly into video nodes without downloading, re-uploading, or losing prompt history. - **First & last frame workflows**: Use start-frame and end-frame inputs on supported cloud video models to guide shot transitions. - **Side-by-side take review**: Evaluate motion stability, framing, and brand details across multiple takes right next to the reference keyframe. - **Audio & storyboard pairing**: Keep voice/music nodes, shot notes, and video clips organized together as a visual sequence. ## Workflow comparisons (https://blendcanvas.com/compare/) ### BlendCanvas vs Local ComfyUI — Cloud Infinite Canvas Without GPU Setup (https://blendcanvas.com/compare/comfyui-cloud-alternative/) Choose BlendCanvas when you want visual reference wiring, multi-model AI image and video comparison, mask editing, and an AI Canvas Agent directly in your browser without local GPUs or personal API keys. Choose local ComfyUI when you need offline execution, custom Python nodes, or low-level tensor and sampler engineering. - **Hardware & setup** — *Alternative*: Requires a dedicated local GPU with sufficient VRAM, model weight downloads, Python dependencies, and custom-node maintenance. | *BlendCanvas*: Opens in any modern browser. AI generation runs on managed cloud channels with zero local GPU or driver setup. - **Model switching & billing** — *Alternative*: Switch checkpoints by downloading multi-gigabyte files or managing separate provider API keys. | *BlendCanvas*: Compare enabled cloud image, video, audio, and text models side by side using one shared account credit balance. - **Graph abstraction** — *Alternative*: Low-level KSampler, VAE, CLIP, and conditioning wires designed for pipeline engineers. | *BlendCanvas*: Creative-first text, reference image, video, and audio nodes built for art directors, designers, and storytellers. - **AI graph building** — *Alternative*: Manual node wiring or importing community JSON graphs that often break on missing nodes. | *BlendCanvas*: Built-in cloud Canvas Agent lays out structured storyboards, character sheets, and connected idle nodes from one brief. - **Storage & external tools** — *Alternative*: Saved locally on one machine unless you self-host storage and servers. | *BlendCanvas*: Projects sync to MySQL and cloud object storage (S3/R2), with authenticated Streamable HTTP MCP access for Claude Code, Codex, and Cursor. ### Infinite Canvas vs Chat-Based AI Image Generators (https://blendcanvas.com/compare/infinite-canvas-vs-chat-generators/) Linear chat generators work well for quick one-off images, but context scrolls out of view as soon as you iterate on a campaign. An infinite canvas keeps product photos, character sheets, style references, and parallel model variations side by side so you can art-direct a cohesive visual series. - **Reference persistence** — *Alternative*: Reference photos get buried in scroll history and must be re-attached or re-explained across turns. | *BlendCanvas*: Reference image and brief nodes stay pinned and wired to every downstream generation branch. - **Multi-model comparison** — *Alternative*: Locked to a single model per thread or split across disconnected browser tabs. | *BlendCanvas*: Wire one prompt and reference node into multiple cloud image models simultaneously and compare outputs at a glance. - **Localized editing** — *Alternative*: Conversational follow-ups frequently alter facial features, product geometry, or lighting unintentionally. | *BlendCanvas*: Select any canvas image node to run localized mask inpainting, aspect-ratio cropping, grid splitting, or high-res upscaling. - **Perspective control** — *Alternative*: Camera angles rely on text guesswork like "three-quarter low angle." | *BlendCanvas*: Preview framing with built-in 3D director and camera angle controls before submitting a generation. - **Image-to-video handoff** — *Alternative*: Download still images manually and upload them into a separate video app. | *BlendCanvas*: Connect approved canvas image nodes directly into cloud image-to-video and storyboard nodes on the same board. ### Connected AI Storyboard Canvas vs Standalone Video Generators (https://blendcanvas.com/compare/multi-model-ai-video-storyboard/) Standalone AI video feeds treat every prompt as an isolated clip, making it expensive to fix lighting or character drift after motion is rendered. Building video on an infinite canvas lets you approve still keyframes first, wire them into cloud video models, and review multi-shot sequences spatially. - **Pre-visualization cost** — *Alternative*: Text-to-video errors waste higher per-second video credits when composition or wardrobe is wrong. | *BlendCanvas*: Iterate composition, mask edits, and camera angles on lower-cost still image nodes before wiring into video. - **Continuity & framing** — *Alternative*: Clips sit in a chronological download list without visual shot order. | *BlendCanvas*: Arrange wide, medium, and close-up keyframes horizontally on the canvas with linked motion prompts and takes. - **Start & end frame wiring** — *Alternative*: Uploading separate start/end files manually for each transition. | *BlendCanvas*: Connect two canvas image nodes directly into compatible first/last-frame cloud video models. - **Model flexibility** — *Alternative*: Tied to a single video engine’s motion style and aspect ratios. | *BlendCanvas*: Choose from enabled cloud video models (such as Veo 3.1, Seedance 2.0, or Wan 3.0) per shot from one credit balance. - **Audio & script context** — *Alternative*: Voice notes, music cues, and shot scripts live in external docs. | *BlendCanvas*: Keep script text nodes, AI music generations, reference stills, and video takes together in one project. ## Canvas workflow templates (https://blendcanvas.com/templates/) ### E-commerce product hero & lifestyle series (https://blendcanvas.com/templates/product-hero-campaign/) Start from one clean product reference and branch into studio hero, macro texture, and sunlit lifestyle shots on a single board. - Inputs: Product reference photo, Brand palette & lighting note - Canvas tools: Reference wiring, Multi-model image nodes, Mask inpainting, Upscale - Starter brief: "Build a 3-branch e-commerce product photography board for a minimalist glass skincare bottle: 1) crisp warm studio hero shot, 2) macro water-droplet texture close-up, and 3) sunlit stone vanity lifestyle scene while keeping bottle silhouette consistent." ### Multi-model AI image style comparison board (https://blendcanvas.com/templates/multi-model-style-shootout/) Wire one creative brief and color reference into three parallel image nodes to compare rendering style, lighting, and typography fidelity. - Inputs: Shared creative brief node, Color & material mood reference - Canvas tools: Parallel model branching, Side-by-side comparison, Crop & upscale - Starter brief: "Create a parallel style comparison board for an independent specialty coffee brand: connect one citrus-and-cobalt palette reference into three distinct visual directions (editorial still life, tactile packaging mockup, and bold poster key visual)." ### Character consistency & multi-angle turnaround sheet (https://blendcanvas.com/templates/character-sheet-turnaround/) Design a base character concept, split turnaround views, and lock costume and silhouette references for downstream story scenes. - Inputs: Character bio & wardrobe prompt, Primary character reference - Canvas tools: Canvas Agent character skill, 3D camera & angle control, Grid split - Starter brief: "Create a character design sheet and 3-angle turnaround board (front, three-quarter, profile) for a stylized sci-fi botanical explorer, keeping jacket silhouette, satchel details, and warm editorial lighting consistent across nodes." ### Keyframe-to-video cinematic storyboard (https://blendcanvas.com/templates/canyon-video-storyboard/) Approve still keyframes on the canvas first, then wire each frame into image-to-video nodes with explicit camera motion prompts. - Inputs: 3 approved scene keyframes, Camera movement notes - Canvas tools: Canvas Agent storyboard skill, Image-to-video nodes, Take comparison - Starter brief: "Lay out a 3-shot cinematic storyboard set in a sunlit sandstone canyon: establishing wide ridge shot, medium tracking shot with drifting dust in warm rim light, and close-up material detail ready for image-to-video generation." ### Industrial & furniture material lighting board (https://blendcanvas.com/templates/architectural-material-study/) Explore brushed chrome, smoked glass, and travertine material swaps from a single form reference without losing earlier iterations. - Inputs: Form silhouette reference, Material & studio lighting prompts - Canvas tools: Single-variable branching, Mask edit, High-res upscale - Starter brief: "Explore a sculptural lounge chair and side table concept across three material studies on one canvas: brushed aluminum with frosted glass, warm walnut with bouclé fabric, and matte terracotta in soft gallery daylight." ## Model workflow spotlights (https://blendcanvas.com/models/) ### GPT Image 2 on an Infinite Canvas — Reference Wiring & Mask Editing (https://blendcanvas.com/models/gpt-image-2/) GPT Image 2 pairs strong prompt adherence and layout structure with BlendCanvas’s spatial reference nodes—ideal for e-commerce hero shots, packaging concepts, and localized mask refinements. - **Multi-reference prompt adherence**: Wire product photos and color references into GPT Image 2 nodes to maintain clear spatial composition across campaign branches. - **Flexible resolution tiers**: Draft compositions at 1K to lock lighting and geometry economically, then switch tiers or upscale approved frames for final delivery. - **Canvas mask inpainting synergy**: Combine structured initial generations with targeted mask editing on the board to adjust props, reflections, or backgrounds. ### Nano Banana on an Infinite Canvas — Rapid Style & Character Iteration (https://blendcanvas.com/models/nano-banana/) Nano Banana excels at brisk creative iteration, moodboard branching, and character concept exploration when connected to visual references and 3D angle cues on an infinite canvas. - **Fast multi-branch moodboarding**: Spin up three lighting or palette directions from a single creative brief without losing track of earlier prompts. - **Character & concept exploration**: Generate multi-angle character studies, split grid views into standalone reference nodes, and wire them into story beats. - **Side-by-side shootout ready**: Pair Nano Banana alongside GPT Image 2 on one board to balance expressive textures with structured commercial layouts. ### Veo 3.1 on an Infinite Canvas — Cinematic Image-to-Video Storyboards (https://blendcanvas.com/models/veo-3-1/) Running Veo 3.1 inside BlendCanvas lets you perfect framing, lighting, and subject details on still image nodes first, then wire approved keyframes straight into cinematic video generation. - **Direct canvas keyframe wiring**: Connect any generated or uploaded canvas image node into Veo 3.1 without exporting files or losing prompt context. - **Cinematic camera & physics cues**: Pair clean visual anchors with explicit camera push, pan, and atmospheric motion prompts on your storyboard. - **Transparent credit & refund handling**: Review exact video credit costs in the workspace before submitting; failed upstream tasks automatically refund reserved credits. ### Seedance 2.0 & Wan 3.0 on an Infinite Canvas — Storyboard Video Workflows (https://blendcanvas.com/models/seedance-and-wan-video/) Seedance 2.0 (Standard, Fast, and Mini tiers) and Wan 3.0 give storyboard creators flexible speed-to-cost choices for animating product reveals, character beats, and environment transitions on one canvas. - **Speed vs fidelity tier selection**: Draft motion blocking quickly with lighter video tiers, then render hero shots at higher specifications on the same board. - **Start & end frame transitions**: On compatible video workflows, wire both a starting and ending canvas image node to guide smooth shot-to-shot motion. - **Multi-shot spatial review**: Keep script notes, still keyframes, Seedance/Wan video takes, and AI music tracks organized in one cloud-synced project. ## Models and pricing - [Model specifications and credit rates](https://blendcanvas.com/models/) - [Subscriptions and credit packs](https://blendcanvas.com/pricing/) - [Live public catalog](https://app.blendcanvas.com/api/catalog): Read-only JSON for enabled models, rates, payment availability and system skills; no sign-in required. - Generation uses the account's shared credit balance. Model, specification and billing unit determine the cost. Failed upstream generations refund reserved credits. ### Enabled model snapshot (verify live availability via https://app.blendcanvas.com/api/catalog) - **Image models**: `gpt-image-2`, `nano-banana`, `gpt-image-2.5-sunburst`, `gemini-3.1-flash-image-preview`, `gemini-nano-banana-2.1`, `gpt-image-2.5-flare`, `gpt-image-2.5`, `gemini-3-pro-image-preview`, `gemini-2.5-flash-image-vip`, `gemini-3-pro-image`, `gpt-image-2-vip`, `gpt-image-2.5-vip`, `gemini-3.1-flash-image`, `nano-banana-pro`, `nano-banana-pro-vip`, `nano-banana-vip`, `gemini-3-pro-image-preview-vip`, `gpt-image-1.5`, `gemini-2.5-flash-image-preview-hd`, `gemini-2.5-flash-image-hd`, `gemini-2.5-flash-image-preview`, `gpt-image-1-vip`, `nano-banana-hd`, `gemini-2.5-flash-image-preview-vip`, `nano-banana-2` - **Video models**: `wan3.0-video`, `veo3.1`, `doubao-seedance-2-0-260128`, `doubao-seedance-2-0-fast-260128`, `doubao-seedance-2.0-mini`, `MiniMax-H3` - **Text models**: `gpt-5.2-all`, `gpt-4o-all`, `o3-mini-all`, `o4-mini-all` - **Audio models**: Suno v6 (`suno-v6`), Suno v6 Wild (`suno-v6-wild`), Suno v6 Mini (`suno-v6-mini`), Suno Auto (`suno-auto`), Suno Lyrics (`suno-lyrics`) ## Remote MCP and developer resources - [MCP setup and supported tools](https://blendcanvas.com/mcp/) - [Public API reference](https://blendcanvas.com/api/) - Remote MCP endpoint: https://app.blendcanvas.com/api/mcp; requires a personal Bearer token. Supported canvas tools: list_canvases, get_canvas_summary, create_canvas, create_text_node. This is not an unauthenticated image or video generation API. ## Workflow guides ### How to compare AI image models and refine details on an infinite canvas (https://blendcanvas.com/guides/multi-model-ai-image-canvas/) Published: 2026-11-18 | Updated: 2026-11-18 To compare AI image models effectively, wire a single reference image and prompt node into multiple image generation nodes on an infinite canvas, evaluate composition and fidelity side by side, then refine the best output with mask inpainting and upscaling. #### Anchor your prompt and reference once Create one text node for your core brief and place your key reference image—such as a product silhouette or lighting study—right beside it. Keeping a single source node prevents prompt drift when you test multiple models. #### Branch into two or three cloud image models Connect the same brief and reference into parallel image nodes, selecting a different enabled cloud model or resolution tier on each node. Some models excel at clean typography and studio packaging, while others produce richer painterly textures. Check each node’s credit rate before running the batch. #### Evaluate structure before chasing small fixes Compare composition, lighting direction, hands or packaging geometry, and color accuracy side by side on the canvas. Pick the strongest structural frame instead of re-rolling a full prompt dozens of times in a chat feed. #### Finish with mask inpainting, angle tools, and upscaling Once a winning composition is selected, use canvas mask editing to fix localized details, split multi-view grids into separate nodes, or run upscaling for final export resolution while keeping earlier versions intact on the board. **Why run multiple image models on the same prompt?** Different AI image models interpret lighting, text rendering, realism, and stylization differently. Testing two or three models side by side on one canvas helps you spot the strongest visual direction quickly without losing your reference context. **When should I use mask inpainting instead of generating a new image?** Use mask inpainting when the overall composition, lighting, and pose already work and only a localized detail—such as a background prop, surface reflection, or accessory—needs adjustment. Always verify critical brand text against your original asset. - Reference: [OpenAI Platform — Image generation and editing guide](https://platform.openai.com/docs/guides/image-generation) - Reference: [Google AI for Developers — Imagen and Gemini image generation](https://ai.google.dev/gemini-api/docs/imagen) ### How to keep character consistency and build storyboards on an AI canvas (https://blendcanvas.com/guides/ai-character-and-storyboard-canvas/) Published: 2026-11-18 | Updated: 2026-11-18 Maintain character consistency on an AI canvas by creating a neutral character reference sheet first, splitting key angles into dedicated reference nodes, wiring those references into each storyboard shot alongside explicit camera framing, and reviewing the sequence side by side. #### Establish a canonical character sheet first Before generating story scenes, create a clean front and three-quarter reference of your character against a neutral background. Record wardrobe colors, props, and proportions in a linked text node, or use the Canvas Agent character-sheet skill to structure the setup. #### Split turnaround views into reusable reference nodes When a generation produces a multi-angle sheet, use the canvas split tool to break individual poses into standalone image nodes. Label each angle clearly so you can wire the closest matching pose into each scene shot. #### Guide shot framing with camera angle controls Plan wide establishing shots, over-the-shoulder dialogue frames, and close-ups as ordered columns on the canvas. Use the 3D camera and angle controls to communicate perspective shifts instead of relying on vague text descriptions alone. #### Bridge approved storyboard frames into video Review the full sequence horizontally across the canvas to check costume, lighting, and color continuity. Connect approved keyframes into compatible image-to-video nodes when you are ready to animate motion. **Does reference wiring guarantee 100% identical characters in every shot?** No AI model guarantees pixel-identical features across every angle. Locking a clean character turnaround reference, changing one scene variable at a time, and correcting minor drift with mask editing significantly improves shot-to-shot continuity. **Can the Canvas Agent lay out a storyboard automatically?** Yes. You can give the cloud Canvas Agent a short story or campaign outline, and it can create structured script notes and connected storyboard nodes on your canvas for review before you spend credits on final renders. - Reference: [Google AI for Developers — Multimodal prompting best practices](https://ai.google.dev/gemini-api/docs/prompting-strategies) - Reference: [Model Context Protocol — Specification & architecture](https://modelcontextprotocol.io/introduction) ### How to plan an AI product photography series (https://blendcanvas.com/guides/product-image-series/) Published: 2026-11-06 | Updated: 2026-11-18 Build an AI product image series from a clear product reference and a consistent visual brief. Generate separate hero, detail, and lifestyle compositions, then check each result against the actual product before publishing. #### Start with a reliable product reference Use a clear image of the actual product. Record the details that must remain consistent, such as silhouette, packaging, color, and label placement. An image model may change these details, so reference input is a starting point rather than a guarantee. #### Choose a visual system For a bottle campaign, plan three shots: a centered hero, a close crop of the cap, and a wider shot on a bathroom shelf. Keep the same warm background palette and soft light from the left. Write these shared choices in a text node, then branch the three shot prompts from the same product reference. #### Generate a small set of variations Select a cloud image model in BlendCanvas, connect your references, and describe one scene at a time. Keep a clean hero composition, a closer detail, and an environmental scene as distinct branches of the project. #### Review before publishing Compare each result to your actual product, especially text, proportions, and materials. Correct or discard inaccurate results. Export only the images that meet your requirements, and check applicable terms for the intended use. **Can AI preserve product labels and packaging exactly?** Exact preservation is not guaranteed. Compare labels, logos, proportions, color, and material with the original product photo. Correct inaccurate details with supported editing tools or reject the image before it reaches a product listing. **How do I keep a product image series consistent?** Use the same product reference and repeat the lighting, background, palette, framing, and aspect ratio in your brief. Change one variable per branch and compare the results side by side. Keep settings with the selected image for later work. **Which image model should I choose for product photography?** Choose an available model that accepts the inputs and editing task you need, then review its resolution options and credit rate. Test it with your actual product before producing a larger set; the concept artwork on this site is not a model benchmark. - Reference: [OpenAI Platform — Image generation and editing guide](https://platform.openai.com/docs/guides/image-generation) - Reference: [Google AI for Developers — Imagen and Gemini image generation](https://ai.google.dev/gemini-api/docs/imagen) ### How to turn an image into a video with AI (https://blendcanvas.com/guides/image-to-video-workflow/) Published: 2026-11-12 | Updated: 2026-11-15 To turn an image into video, upload a clear source frame, choose a compatible image-to-video model, describe the motion, review its settings and credit cost, then compare the result with your original frame. #### Choose a source image that can carry the shot Start with a clear subject and a composition close to your intended shot. Leave room in the direction the subject or camera will move. For a skincare product clip, try a single bottle on a neutral surface with its label already facing the camera. Add the image to your canvas and keep the original nearby for review. #### Connect the image to a compatible video model Add a video node and connect your image node as an input. Choose a cloud model that accepts an image reference. If the model offers a first-frame control and you want the clip to start with your image, set it there. Check the model settings for support before planning a shot around an end frame or multiple references. #### Describe the motion and what should stay fixed Name the subject, camera movement, and any details that should stay stable. Try this product-shot prompt: “The camera slowly pushes in toward the bottle. The bottle stays still, its front label faces the camera, and a soft shadow moves across the table.” Start with one camera move so you can identify what changed when you compare takes. #### Generate a first take at the intended aspect ratio Choose the aspect ratio and duration from the settings the model exposes. A vertical social clip and a landscape product banner need different framing, so make that choice before generating. Check the credit cost for the selected settings and generate a take with your account credits. #### Review the clip against the source and refine one choice Compare the clip with the source image on the canvas. Check the label, silhouette, framing, and motion near the start and end of the clip. If the camera moves too fast or the product drifts, keep the first take and change one instruction for the next. If the model distorts a detail you need to preserve, adjust the input or discard the take before publishing. **What should I include in an image-to-video prompt?** Describe what moves, how it moves, the camera movement, and what should stay consistent. For example, ask for a slow camera push toward a product with gentle background motion while keeping the packaging shape unchanged. Review the output because a prompt does not guarantee consistency. **Can I control the first and last frame?** Use first- and last-frame inputs only when the selected model and workflow support them. Available inputs, duration, resolution, and credit costs vary by model; check the workspace settings before generating. **How can I compare different video takes?** Keep the source image and motion prompt on the canvas, generate a small number of takes, and compare subject shape, motion, camera framing, and unwanted changes. Refine one part of the prompt at a time. - Reference: [Google AI for Developers — Video generation with Veo](https://ai.google.dev/gemini-api/docs/video) ### From a loose brief to a focused creative board (https://blendcanvas.com/guides/creative-brief-to-board/) Published: 2026-11-06 | Updated: 2026-11-18 Turn a brief into a creative board by recording the audience, deliverables, and constraints, adding a small set of labeled references, and branching one visual choice at a time. Keep the selected prompts and results together for the next iteration. #### Define the outcome first Write down the audience, the asset you need, its format, and where it will appear. A product hero image and a social campaign need different compositions. Keep these constraints in a text node so they stay visible throughout the project. #### Build a small reference board Add a few references that each communicate a specific choice: lighting, color, composition, or material. Use images you own or have permission to use. Label what matters in each reference rather than collecting a large, unfocused mood board. #### Explore one variable at a time Start with one direction. Change the background, framing, or lighting separately so you can understand what improves the result. Keep the original brief and earlier versions next to each variation on the canvas. #### Save a repeatable workflow Keep the useful prompt, reference images, model settings, and selected output together. Duplicate the canvas for the next project and replace the inputs. Results depend on your selected model; check details before using an output in a finished campaign. **What belongs in a creative brief?** Record the audience, the purpose of the asset, required formats, where it will appear, and details that must stay consistent. Add references that explain lighting, color, or composition rather than collecting images without a clear reason. **How do I reuse a creative board?** Duplicate a useful canvas, replace the input references and brief, and review the connected prompts and settings before generating. A copied board preserves a starting structure; it does not guarantee identical results with a new product or model. - Reference: [Google AI for Developers — Multimodal prompting best practices](https://ai.google.dev/gemini-api/docs/prompting-strategies) - Reference: [OpenAI Platform — Prompt engineering & visual workflows](https://platform.openai.com/docs/guides/prompt-engineering) ## Frequently asked questions ### What is BlendCanvas? BlendCanvas is an AI image generator and creative workspace built around an infinite canvas. Instead of losing earlier prompts in a linear chat, you connect reference images, text briefs, AI image variations, video clips, and audio nodes on one visual board. ### How is BlendCanvas different from chat-based AI image generators or ComfyUI? Linear chat tools bury earlier references and make side-by-side comparison slow, while local node tools like ComfyUI require GPU hardware, model downloads, and complex operator graphs. BlendCanvas gives you visual reference wiring, multi-model comparison, mask inpainting, upscaling, and camera angle controls directly in the browser with managed cloud models. ### Do I need to configure an AI provider? No. BlendCanvas provides managed cloud models directly in the workspace—you do not need to bring your own API keys or configure local endpoints. Simply choose a model and specification, and generations use your account credits based on the published credit rate. ### Can I start without an account? Signing in is required to use the workspace. Your canvases and account data are saved in the cloud database, media files are stored in cloud object storage, and your browser keeps an account-isolated cache for fast loading. ### Can I use my own images and references? Yes. Add your images, text, video, and audio to your canvas and use compatible inputs with the model you select. Use materials you own or have permission to use. ### Are the images on this page generated in BlendCanvas? These are AI-generated concept artworks used to illustrate creative directions. They are not customer projects, testimonials, or benchmark results from the workspace. ## Company and content provenance - Operator: Quguo Limited, Hong Kong - Public contact: support@blendcanvas.com - [Company and contact](https://blendcanvas.com/about/) - Concept artwork illustrates creative directions. It is not a customer testimonial, a verified workspace result, or evidence of model performance. - [Terms of Service](https://blendcanvas.com/legal/terms/) - [Service Agreement](https://blendcanvas.com/legal/service-agreement/) - [Privacy Policy](https://blendcanvas.com/legal/privacy/) - [Disclaimer](https://blendcanvas.com/legal/disclaimer/) - [Cookie & Browser Storage Policy](https://blendcanvas.com/legal/cookies/) - [Refund Policy](https://blendcanvas.com/legal/refund/) ## 简体中文参考(融绘 BlendCanvas) > 融绘 BlendCanvas 的官网是 blendcanvas.com。在这款 AI 图像与视频工作台中,你可以连接产品参考图、对比多版图像,并把不同的视频片段放在同一张画布上。 ### 核心能力 - **参考图与提示词节点连线**:将产品实拍图、布光参考、角色设定图与文本简报直接连线至图片或视频生成节点,让创作上下文全程可视。 - **多模型生图同屏横评**:将同一提示词或参考图分支连接到多款云端图像模型与不同分辨率档位,直观对比构图、文字表现与材质还原度。 - **局部蒙版重绘与智能放大**:无需离开画布即可使用蒙版局部重绘修改细节、对选中画面进行高清放大,或对多宫格出图进行拆分与裁剪。 - **3D 导演台与机位视角控制**:借助内置 3D 导演台与视角控制工具预览主体构图和镜头透视,辅助生成视角一致的多机位画面。 - **云端画布智能助手(Agent)**:只需输入简短需求,云端画布助手即可自动编排角色设定表、分镜故事板序列或多分支情绪板节点图。 - **远程 MCP 协议接入**:通过个人令牌与 Streamable HTTP 连接 Claude Code、Codex、Cursor 等 MCP 客户端,直接读取与起草画布节点。 ### 功能模块 ### AI 生图无限画布(https://blendcanvas.com/zh/features/ai-image-workspace/) 在无限画布上生成、对比并精修 AI 图片。连接产品图或风格参考,向多款云端生图模型并行分支,并结合局部蒙版重绘、机位视角控制与高清放大打磨成片。 - **多参考图可视化连线**:让产品轮廓、品牌色板与构图参考始终与每一张生成变体保持清晰连线。 - **多模型同场出图对比**:在同一画布上用多款云端生图模型测试同一提示词,无需来回切换标签页或重复上传参考素材。 - **局部重绘、放大与多图拆分**:涂抹蒙版进行定向局部重绘,将九宫格/多视图拆分为独立图片节点,自由裁剪构图并高清放大定稿画面。 - **3D 导演台与镜头角度**:在 3D 人台与机位舞台上预览镜头角度,为产品主图、特写与多角度视图提供一致的透视参考。 ### 面向 AI 创意工作流的无限画布(https://blendcanvas.com/zh/features/infinite-canvas/) 在云端同步的无限画布上统一组织创意简报、参考板、AI 图片、视频片段与音频——内置云端画布 Agent 并支持远程 MCP,无需配置本地显卡或折腾 ComfyUI。 - **无损视觉分支探索**:在并行分支中试验不同布光、文案与构图方案,始终保留原始参考素材不被覆盖。 - **画布 Agent 自动建图**:输入一句自然语言需求,即可让云端助手起草剧本、角色设定、分镜节点与连线布局。 - **免本地显卡与免 API Key 配置**:直接在浏览器中调用托管云端模型运行多步视觉工作流,共用同一账号积分余额。 - **云端同步与托管 MCP**:画布项目同步至 MySQL 与云端对象存储,并提供鉴权远程 MCP 工具供外部 AI 客户端连接。 ### AI 图生视频与分镜故事板工作空间(https://blendcanvas.com/zh/features/ai-video-workspace/) 将画布生成的静帧或上传的参考图转化为 AI 视频片段。在融绘 BlendCanvas 画布上并排保留源关键帧、运镜提示词、首尾帧控制与多条成片镜头。 - **生图到视频无缝衔接**:将画布上定稿的图片节点直接连入视频节点,无需反复下载重传,完整保留提示词上下文。 - **首尾帧引导控制**:在支持的云端视频模型上接入首帧与尾帧图片,稳定控制镜头起止过渡。 - **多条镜头同屏审片**:紧挨参考关键帧横向审视多条生成镜头的运动稳定性、构图与品牌细节。 - **配乐音频与分镜编排**:将人声/音乐节点、镜头脚本批注与视频片段按分镜顺序组织在同一空间。 ### 方案对比、画布模板与创作指南 - [工作流方案对比](https://blendcanvas.com/zh/compare/) - [AI 画布生图模板库](https://blendcanvas.com/zh/templates/) - [AI 模型目录](https://blendcanvas.com/zh/models/) - [订阅与积分定价](https://blendcanvas.com/zh/pricing/) - [云端无限画布 vs 本地 ComfyUI](https://blendcanvas.com/zh/compare/comfyui-cloud-alternative/):如果你希望在浏览器中直接使用参考图连线、多模型生图与视频同屏对比、局部蒙版重绘及 AI 画布 Agent,而不想维护本地显卡或个人 API Key,请选择 BlendCanvas;如果你需要离线断网运行、编写自定义 Python 算子节点或底层采样器调优,本地 ComfyUI 更合适。 - [无限画布 vs 线性聊天生图](https://blendcanvas.com/zh/compare/infinite-canvas-vs-chat-generators/):线性对话工具适合随手生成单张草图,但一旦进入多轮企划迭代,早期参考图很快就会淹没在滚动记录里。无限画布将产品原图、角色设定、风格参考与多模型并行变体平铺在同一空间,方便你把控整套视觉作品的一致性。 - [连线分镜画布 vs 独立视频流](https://blendcanvas.com/zh/compare/multi-model-ai-video-storyboard/):独立视频生成器把每次提交当作孤立短片,一旦渲染后发现光影或角色走形,返工成本极高。在无限画布上先定稿低成本静态关键帧,再连线送入云端视频模型并横向排布多镜头成片,能让分镜叙事更可控。 - [如何在无限画布上对比多款 AI 生图模型并精修局部细节](https://blendcanvas.com/zh/guides/multi-model-ai-image-canvas/):高效对比 AI 生图模型的方法是:在无限画布上将同一参考图与提示词节点连线至多个生图节点,并排评估构图与还原度,再对最佳结果执行局部蒙版重绘与高清放大。 - [如何在 AI 画布上保持角色一致性并制作分镜故事板](https://blendcanvas.com/zh/guides/ai-character-and-storyboard-canvas/):在 AI 画布上保持角色一致性的核心流程是:先制作中性背景的角色基准设定图,拆分关键朝向作为独立参考节点,按每个分镜的景别机位连线对应参考,并在同一画布横向核对全片连贯性。 - [如何规划一组 AI 产品摄影图](https://blendcanvas.com/zh/guides/product-image-series/):从清晰的产品参考图和统一的视觉简报开始,分别生成主图、细节图与场景图,再逐张对照真实产品检查后发布。 - [如何用 AI 把图片变成视频](https://blendcanvas.com/zh/guides/image-to-video-workflow/):将图片转为视频时,先上传清晰的原始画面,选择兼容的图生视频模型并描述运动,确认设置与积分费用,再将结果与原图对照。 - [从模糊需求到清晰的创意板](https://blendcanvas.com/zh/guides/creative-brief-to-board/):先记录受众、交付物和限制条件,再添加少量带说明的参考素材,每次围绕一个视觉选择创建分支。将选定的提示词和结果放在一起,便于下一轮迭代。 ### 常见问题(中文) #### 融绘 BlendCanvas 是什么? 融绘 BlendCanvas 是围绕无限画布构建的 AI 生图与多模态创作工作空间。告别在冗长聊天记录里翻找早期提示词,你可以在同一块可视画板上连接参考图、文本简报、AI 生图分支、视频片段与音频节点。 #### 融绘 BlendCanvas 与对话式 AI 生图工具或 ComfyUI 有什么区别? 线性聊天生图容易淹没早期参考图、难以横向对比;而 ComfyUI 等本地节点工具需要配置显卡算力、下载模型权重并搭建复杂算子图。融绘 BlendCanvas 在浏览器内即开即用,直接结合托管云端模型提供可视化参考连线、多模型对比、局部重绘、高清放大与镜头视角控制。 #### 需要配置 AI 模型服务商吗? 不需要。融绘 BlendCanvas 在工作台内直接提供托管云端模型——你无需自备 API 密钥或配置本地接口。只需选择模型与规格,生成任务即按公开积分费率使用账号积分结算。 #### 不注册也能开始吗? 使用工作台需要登录账号。你的画布与账号数据保存在云端数据库,媒体文件存储在云端对象存储,浏览器保留按账号隔离的本地缓存以加快加载速度。 #### 可以使用自己的图片和参考吗? 可以。将图片、文本、视频和音频加入画布,并配合所选模型支持的输入使用。请使用你拥有或已获授权的素材。 #### 页面上的图片由 融绘 BlendCanvas 生成吗? 这些图片为 AI 生成的创意概念图,用于展示视觉方向,不代表客户项目、评价或工作台的性能测试结果。 ## 繁體中文參考(融繪 BlendCanvas) (https://blendcanvas.com/zh-tw/) > 融繪 BlendCanvas 的官網是 blendcanvas.com。在這個 AI 圖像與影片工作台中,你可以連接產品參考圖、比較不同版本的圖像,並將多個影片片段放在同一張畫布上。 - **參考圖與提示詞節點連線**: 將產品實拍圖、布光參考、角色設定圖與文本簡報直接連線至圖片或視頻生成節點,讓創作上下文全程可視。 - **多模型生圖同屏橫評**: 將同一提示詞或參考圖分支連接到多款雲端圖像模型與不同分辨率檔位,直觀對比構圖、文字表現與材質還原度。 - **局部蒙版重繪與智能放大**: 無需離開畫布即可使用蒙版局部重繪修改細節、對選中畫面進行高清放大,或對多宮格出圖進行拆分與裁剪。 - **3D 導演台與機位視角控制**: 借助內置 3D 導演台與視角控制工具預覽主體構圖和鏡頭透視,輔助生成視角一致的多機位畫面。 - **雲端畫布智能助手(Agent)**: 只需輸入簡短需求,雲端畫布助手即可自動編排角色設定表、分鏡故事板序列或多分支情緒板節點圖。 - **遠程 MCP 協議接入**: 通過個人令牌與 Streamable HTTP 連接 Claude Code、Codex、Cursor 等 MCP 客戶端,直接讀取與起草畫布節點。 - [畫布生圖模板](https://blendcanvas.com/zh-tw/templates/) · [對比](https://blendcanvas.com/zh-tw/compare/) · [模型列表](https://blendcanvas.com/zh-tw/models/) · [定價](https://blendcanvas.com/zh-tw/pricing/) - [雲端無限畫布 vs 本機 ComfyUI](https://blendcanvas.com/zh-tw/compare/comfyui-cloud-alternative/): 如果你希望在瀏覽器中直接使用參考圖連線、多模型生圖與影片同螢幕對比、局部遮罩重繪及 AI 畫布 Agent,而不想維護本機顯卡或個人 API Key,請選擇 BlendCanvas;如果你需要離線斷網執行、編寫自訂 Python 算子節點或底層採樣器調校,本機 ComfyUI 更合適。 - [無限畫布 vs 線性聊天生圖](https://blendcanvas.com/zh-tw/compare/infinite-canvas-vs-chat-generators/): 線性對話工具適合隨手生成單張草圖,但一旦進入多輪企劃迭代,早期參考圖很快就會淹沒在捲動記錄裡。無限畫布將產品原圖、角色設定、風格參考與多模型並行變體平鋪在同一空間,方便你把控整套視覺作品的一致性。 - [連線分鏡畫布 vs 獨立影片流](https://blendcanvas.com/zh-tw/compare/multi-model-ai-video-storyboard/): 獨立影片生成器把每次提交當作孤立短片,一旦渲染後發現光影或角色走形,返工成本極高。在無限畫布上先定稿低成本靜態關鍵影格,再連線送入雲端影片模型並橫向排佈多鏡頭成片,能讓分鏡敘事更可控。 ## Référence en français (BlendCanvas) (https://blendcanvas.com/fr/) > BlendCanvas est l’espace de création d’images et de vidéos par IA disponible sur blendcanvas.com. Reliez vos références produit, comparez les variantes d’une image et regroupez vos prises vidéo sur le même canevas. - **Références connectées par nœuds**: Reliez directement photos produit, références d'éclairage, fiches personnages et briefs texte aux nœuds de génération d'image ou de vidéo pour garder tout le contexte visible. - **Comparaison d'images multi-modèles**: Déclinez un même prompt ou une même référence sur plusieurs modèles d'image cloud et niveaux de résolution pour comparer côte à côte la composition, la typographie et le rendu des matières. - **Inpainting par masque et agrandissement intelligent**: Affinez des zones précises grâce à la retouche par masque sur le canevas, agrandissez les images sélectionnées en haute résolution, ou recadrez et découpez des générations multi-vignettes sans quitter votre espace. - **Contrôle caméra et angles 3D**: Ajustez le cadrage du sujet et la perspective de la caméra grâce au réalisateur 3D et aux outils d'angle intégrés avant de générer des plans multi-angles cohérents. - **Canvas Agent cloud**: Demandez à l’assistant cloud de transformer un court brief en planche de rotation de personnage, séquence de storyboard ou moodboard à plusieurs branches. - **Intégration MCP distante**: Connectez Claude Code, Codex, Cursor ou d'autres clients MCP via Streamable HTTP avec un jeton personnel pour inspecter et préparer des nœuds sur le canevas. - [Modèles de canevas](https://blendcanvas.com/fr/templates/) · [Comparer](https://blendcanvas.com/fr/compare/) · [Modèles](https://blendcanvas.com/fr/models/) · [Tarifs](https://blendcanvas.com/fr/pricing/) - [Canevas cloud vs ComfyUI local](https://blendcanvas.com/fr/compare/comfyui-cloud-alternative/): Choisissez BlendCanvas si vous souhaitez relier vos références visuelles, comparer plusieurs modèles image et vidéo, retoucher par masque et utiliser un Canvas Agent directement dans le navigateur sans GPU local ni clé API personnelle. Choisissez ComfyUI en local si vous avez besoin d’exécution hors ligne, de nœuds Python sur mesure ou de réglages bas niveau. - [Canevas infini vs Chat linéaire](https://blendcanvas.com/fr/compare/infinite-canvas-vs-chat-generators/): Les générateurs en chat linéaire conviennent à une image isolée, mais le contexte disparaît dès qu’une campagne s’allonge. Un canevas infini garde photos produits, fiches personnages, références de style et variantes multi-modèles côte à côte pour diriger une série cohérente. - [Canevas storyboard vs Flux vidéo](https://blendcanvas.com/fr/compare/multi-model-ai-video-storyboard/): Les flux vidéo isolés traitent chaque prompt séparément, rendant coûteuse toute correction d’éclairage ou de personnage après rendu. Sur un canevas infini, vous validez d’abord les images clés fixes avant de les relier aux modèles vidéo cloud et d’examiner toute la séquence dans l’espace. ## Referencia en español (BlendCanvas) (https://blendcanvas.com/es/) > BlendCanvas es el espacio de trabajo de blendcanvas.com para crear imágenes y vídeos con IA. Conecta referencias de productos, compara versiones de una imagen y reúne tus tomas de vídeo en el mismo lienzo. - **Referencias conectadas por nodos**: Conecta fotos de producto, referencias de iluminación, fichas de personajes y briefs de texto directamente a nodos de generación de imagen o vídeo para mantener el contexto visible. - **Comparación de imágenes multimodelo**: Ramifica un prompt o una referencia en varios modelos de imagen en la nube y niveles de resolución para comparar composición, tipografía y fidelidad de materiales lado a lado. - **Retoque por máscara (inpainting) y escalado inteligente**: Refina regiones específicas con edición por máscara en el lienzo, mejora la resolución de los fotogramas seleccionados o recorta y divide generaciones multipanel sin salir del tablero. - **Control de cámara y ángulos 3D**: Ajusta el encuadre del sujeto y la perspectiva de cámara con el director 3D y las herramientas de ángulo integradas antes de generar tomas multiángulo coherentes. - **Canvas Agent en la nube**: Pide al asistente en la nube que convierta un brief breve en una hoja de giro de personaje estructurada, una secuencia de storyboard o un moodboard de múltiples ramas. - **Integración MCP remota**: Conecta Claude Code, Codex, Cursor u otros clientes MCP mediante Streamable HTTP con un token personal para inspeccionar y preparar nodos del lienzo. - [Plantillas de lienzo](https://blendcanvas.com/es/templates/) · [Comparar](https://blendcanvas.com/es/compare/) · [Modelos](https://blendcanvas.com/es/models/) · [Precios](https://blendcanvas.com/es/pricing/) - [Lienzo en la nube vs ComfyUI local](https://blendcanvas.com/es/compare/comfyui-cloud-alternative/): Elige BlendCanvas cuando quieras conectar referencias visuales, comparar varios modelos de imagen y vídeo, editar con máscaras y usar un Canvas Agent en el navegador sin GPU local ni claves API propias. Elige ComfyUI local cuando necesites ejecución sin conexión, nodos Python personalizados o ingeniería de samplers a bajo nivel. - [Lienzo infinito vs Chat lineal](https://blendcanvas.com/es/compare/infinite-canvas-vs-chat-generators/): Los generadores por chat funcionan bien para imágenes sueltas rápidas, pero el contexto se pierde al desplazarse durante una campaña. Un lienzo infinito mantiene fotos de producto, hojas de personaje, referencias de estilo y ramas multimodelo lado a lado para dirigir una serie visual coherente. - [Lienzo storyboard vs Feeds de vídeo](https://blendcanvas.com/es/compare/multi-model-ai-video-storyboard/): Los generadores de vídeo independientes tratan cada prompt como un clip aislado, encareciendo corregir iluminación o personajes tras renderizar. En un lienzo infinito apruebas primero los fotogramas fijos, los conectas a modelos de vídeo en la nube y revisas la secuencia espacialmente.