| Capability | Description |
|---|
| Research & planning | Web search, social listening, brand analysis, content ideation, model selection |
| Image | Generates from text, edits existing images. Reference images for visual consistency |
| Video | Generates from text, images, or existing video. Multi-shot, lip-sync, start + end frames |
| Audio | Voices across 60+ languages. Multiple music models. Sound effects. Voice transformation. Custom audio uploads. Multi-track mixing |
| Motion graphics | Kinetic typography, data visualizations, logo reveals, custom effects |
| Cost tracking | Estimates cost per operation before execution. Tracks spending and enforces budgets |
| Parallel execution | Up to 10 generations running simultaneously per project |
| Media analysis | Analyzes uploaded and generated media for quality, content, and alignment |
| Post-processing | 4x image upscale, 2x video upscale, background removal, image cropping |
| Agentic editing | Builds and reorders scenes, adjusts durations, syncs audio to visuals, places text/logo/motion overlays, generates captions with word-level timing. Works with generated and uploaded content |
| Variations | Branches the timeline into independent versions for different formats, languages, or creative directions |
| Rendering | Composites all layers into a single MP4 at 30 fps. 1080p and 4K |
| Failure recovery | Automatic retries and recovery. Keeps working through model or provider issues |
| Memory & learning | Per-project memory. Cross-project learned preferences |
| Custom skills | Brand guidelines, style rules, and domain knowledge that persist across projects |
Models
| Model | Best for |
|---|
| Claude Opus 4.6 (default) | Most capable. Complex multi-scene projects, nuanced direction. |
| Claude Sonnet 4.6 | Faster responses, lower cost per run. Straightforward tasks. |
Modes
| Mode | How it works | Default in |
|---|
| Collaborative | Pauses at each stage for review and approval. Plan, cost estimate, generated outputs. | Studio |
| Autonomous | Handles everything end to end without checkpoints. | API |
Memory
Per-project context that builds as you work with the agent. Shared across all variations of the timeline. The agent updates memory automatically and you can see each update in the chat.
| Section | What it stores |
|---|
| Creative | The idea, storyline, visual direction |
| Plan | What’s done, what’s next, notes |
| Assets | URLs, voice IDs, style and character references |
Continuous learning
Eversince picks up your preferences over time and applies them across projects.
| What it learns | |
|---|
| Visual | Styles, color palettes, shot compositions |
| Brand | Recurring elements, tone, messaging |
| Production | Model preferences, workflows, output formats |
Custom skills
Persistent instructions you can add, like brand guidelines, style rules, tone of voice, domain knowledge, or preferred workflows. Applied on every agent run across every project when active.
| Detail | Value |
|---|
| Create | Studio or POST /account/skills via the API |
| Character budget | 8,000 characters shared across all active skills |
| Toggle | Activate or deactivate on the fly without deleting |
| Scope | Applied across all projects where active |
Voice mode
A voice agent that extracts actionable items for Eversince to execute. Available from the prompt box on the home page and in the studio. Male or female voice, persisted across sessions.
| Model | Description |
|---|
| GPT Realtime 1.5 | Better for natural conversation with low latency |
| Claude Sonnet 4.6 | Better for creative reasoning with multimodal context |
Both models have the production knowledge and are environment-aware. Claude Sonnet 4.6 also sees images throughout the voice session.
