Get started in 5 minutes
Complete one small end-to-end run before adding more models, assets or publishing accounts.
Sign in to your workspace
Use an email or sign-in method authorized by your administrator. Ask your team admin if access is missing; never share an account.
Define the outcome
Write down the platform, audience, language, aspect ratio, duration and desired action. A clear outcome matters more than extra adjectives.
Generate one sample
Choose a model currently available in your workspace and start with a short sample. Wait for a final status, then review people, motion, text and sound.
Organize and edit
Save approved media to the asset library. Arrange clips, captions, overlays and audio in the workbench, then export a platform-ready version.
Review, then publish
Confirm account, copy, language, rights, format and timing. Start with a small relevant canary, then review publication status and performance.
Use EMS by task
Each task includes a practical path, a completion check and one commonly missed detail.
Generate video and images
- Open video or image generation and use the models and parameters currently shown.
- Describe subject, action, setting, camera, lighting/style, then sound or caption requirements.
- Only upload reference media you may use and state which elements must remain consistent.
- Create a few candidates first; inspect people, hands, text, branded UI, language and visual stability one by one.
Manage assets and edit online
- Before upload, confirm source, usage rights, language, aspect ratio and file playback.
- Name assets by project, version, language and purpose instead of using ambiguous names such as final-2.
- Select shots before adding captions, overlays and audio; build longer videos from approved short shots.
- Before export, review the opening, transitions, caption safe area, ending and audio sync from start to finish.
Connect accounts and publish
- Use only authorized social accounts that fit the current content and audience.
- Use natural local language, captions and calls to action, and confirm platform format and content rules.
- Choose immediate or scheduled publishing; recheck account, timezone, cover, title and final media before submitting.
- Start with a small canary and expand only after the publication record reaches a final state.
Track status and learn
- Use the shared record after generation or publishing; queued or processing does not mean success.
- Read the failure reason before retrying. Do not repeatedly submit the same failed request.
- Compare views, engagement and conversion while recording creative, language, account and publish time.
- Keep effective templates and stop combinations with poor performance, low quality or safety risk.
Best practices
Reliable output comes from clear goals, small tests and strict quality control—not from generating a large batch at once.
Write prompts like shot directions
Describe who does what, where it happens, camera movement, light and pacing, then sound or caption needs.
Sample before scaling
Begin with one short sample and a few candidates. Increase count, duration or cost only after the direction passes review.
Protect product visuals separately
For a real app, logo, buttons or workflow, inspect every frame for clarity, authenticity and readable text.
Localize the whole creative
Adapt people, setting, spoken language, voice, captions, rhythm and CTA. Caption-only translation is rarely enough.
Use four pre-publish gates
Creative difference, non-product audio/caption difference, authentic product UI, and natural people and motion must all pass.
Preserve provenance and decisions
Record source, rights, version, candidate rationale, target account and result for collaboration, learning and audit.
Troubleshooting
Why can’t I see a model, account or menu?
Availability depends on workspace configuration, role permissions and account scope. Refresh once, then ask your administrator; do not try to bypass the restriction.
Why is a job still queued or processing?
Wait and refresh its status. Processing is neither failure nor success. If it exceeds your team’s expected time, share the timestamp and sanitized error text with support.
Why doesn’t the result follow my prompt?
Shorten the prompt, keep one primary action and camera direction, and remove conflicting requirements. Hold constants steady and change one variable at a time.
What if captions, a logo or phone UI are distorted?
Do not publish. Use a clear authorized reference, separate product visuals from generated human footage, and inspect the final composition frame by frame.
What should I do when publishing fails?
Check account health, authorization, platform rules, file format and timing. Correct the reported cause before retrying; avoid repeated identical submissions.
How do I request access or product support?
Use the email in the footer and include your team, target platform, affected step and time. Never send passwords, codes, cookies, tokens or API keys.
Security and privacy
Public help never displays tenant data. Apply the same minimum-disclosure rule whenever you share a problem or media.
Do
- Use your own authorized identity and the least privilege required.
- Confirm media rights, likeness, trademarks and platform rules before publishing.
- Redact email addresses, account identifiers, business data and internal links from screenshots.
- Stop immediately and notify an administrator after suspicious access, accidental publishing or possible exposure.
Do not
- Do not share passwords, verification codes, cookies, access tokens or API keys.
- Do not place secrets in prompts, source control, public docs, screenshots or videos.
- Do not upload personal, client or confidential media without authorization.
- Do not connect EMS through unknown scripts, plugins or links.
Need automation?
Start with the web workspace. Read the REST API, Agent Skill or MCP documentation only after you are authorized and understand the permission boundary.
Help Center
Learn product workflows, permissions, accounts, publishing and common questions.
Read docs →REST API and OpenAPI
Discover models, accounts, and endpoints before using generation, publishing, personas, analytics, sessions, and remote machines.
View API →EMS Agent Skill
Download an inspectable Skill for Gemini CLI, Cursor, Claude Code and Codex.
Install Skill →Remote MCP
Use 16 EMS tools, resources and an automation prompt over Streamable HTTP.
Configure MCP →Clear rate boundaries for every automation call
REST and remote MCP requests are counted simultaneously by API key, account, and client IP. Reaching any limit stops work before business processing, reducing abuse and runaway cost.
- Read
- 120 · 240 · 300
- Write
- 30 · 60 · 120
- Billable
- 5 · 10 · 20
- Upload
- 300 · 600 · 900
- MCP
- 60 · 120 · 180
Still need help?
Send the steps, timestamp, browser and sanitized error text. Remove all secrets, personal information, tenant accounts and business data first.