// Why I built this
Why I built this
My frustration with AI tools wasn’t that they were AI. Any tool, AI or otherwise, lives or dies on its usefulness in the edit. I have cut shots out of an edit I planned for days and shot for hours. What killed these tools was simpler: they kept pulling me out of Premiere. Export, go somewhere else, wait, and re-import. The round trip killed the momentum.
Temp VO is a good example. Scratch narration sounds fine AI-generated, good enough for a rough cut. But the workflow was: leave Premiere, go to a cloud tool, build the VO, download the file, then figure out what landed where and start reorganizing, housekeeping instead of editing. There was no way to just do it inside Premiere. So I started building a panel that could.
One experiment led to another. That became Asterion. Not a vision for the future of filmmaking, just what happened when a working editor got annoyed enough to start building.
// Asterion Generate
AI generation inside Adobe Premiere Pro for editors.
Asterion Generate is the Premiere Pro workflow layer for model routing, prompts, previews, libraries, and import. Each tab supports a different part of the same post-production loop: source material, generated options, voice, sync, results, and setup.
// Asterion Generate
Generate from the material already in the edit.
Load a selected Premiere clip, attach reference frames, shape the prompt, and keep results in a library instead of another download folder. Selected outputs can be brought back into the Premiere project bin when they are ready.
- Clip, image, and prompt starts
- Missing inserts and temp visuals
- Reusable output library
// Models
Curated routes instead of a random model menu.
Asterion Generate starts with practical fal.ai and Atlas Cloud routes for Seedance, Kling, Wan, LTX, Hailuo, Vidu, Seedream, Imagen, Flux, and Qwen, with Kokoro and ElevenLabs covering voice and sync work.
- fal.ai and Atlas Cloud routes
- Kokoro local voice
- ElevenLabs voice and sync
// Voice
Scratch narration and polished reads without rebuilding the workflow.
Use local voice for private drafts, cloud voice for client-ready options, then keep the audio tied to the same post pipeline.
- Local scratch reads
- Cloud voice options
- Pitch VO and timing tests
// Sync
Prepare lip-sync and voice-clone passes in the same panel family.
Sync is where dialogue replacement, localization tests, and consent-gated clone workflows stay reviewable before anything lands back in the edit.
- Dialogue replacement tests
- Localization passes
- Consent-aware voice workflows
// Library
Keep outputs visible after the model finishes.
Generated shots, voice files, and sync outputs need a place to live. Asterion treats the library as part of the workflow, not an afterthought.
- Review generated assets
- Reuse promising outputs
- Import selected results
// Settings
Setup, bridge state, and production paths stay explicit.
Editors should know what is connected, where files are going, and which workflow settings are active before they run a job.
- Workflow setup
- Bridge health
- Output locations