◇ Integrations
Every external AI goes through a thin adapter. Drop a key into .env.local and the router picks it up — no code change, no redeploy.
default for → long-form, video script, brainstorm, video clip
default for → social, newsletter
default for → image (hero)
default for → voiceover
Which provider runs each modality. Change the dropdown to pin an override — it persists for this workspace and applies on the next generation. Pick "use default" to revert.
| Modality | Active provider | Override |
|---|---|---|
| long-form | Anthropic Claude (Sonnet 4.5) (hero · matrix default) | |
| social | Google Gemini (2.5 Flash) (standard · matrix default) | |
| video script | Anthropic Claude (Sonnet 4.5) (hero · matrix default) | |
| newsletter | Google Gemini (2.5 Flash) (standard · matrix default) | |
| brainstorm | Anthropic Claude (Sonnet 4.5) (hero · matrix default) | |
| image (hero) | Flux (image, via fal) (standard · matrix default) | |
| video clip | Seedance (video, via fal) (hero · matrix default) | |
| voiceover | ElevenLabs (voice, via fal) (standard · matrix default) |
Future-modality adapters. Each lights up when its pipeline ships and the matching env var is set.
Self-hosted open-weights LLM via Ollama. Add OLLAMA_HOST and start the daemon to enable.
cost: $0 (local GPU) · env: OLLAMA_HOST
Meta MusicGen self-hosted for soundtrack + reel beds.
cost: $0 (self-hosted) · env: MUSICGEN_URL
Open-source talking-head avatar engine. Used by UGC + presenter pipelines.
cost: $0 (self-hosted) · env: SADTALKER_URL
OpenAI Whisper self-hosted for transcription + dubbing pipelines.
cost: $0 (self-hosted) · env: WHISPER_URL