The VS Code of video editing, built for agents.
TypeScript
Update model references and package dependencies for improved functionality
konstantin-paulus32ab74e482b2
diff --git a/apps/web/.env.example b/apps/web/.env.exampleindex fac63030..d6d1f780 100644--- a/apps/web/.env.example+++ b/apps/web/.env.example@@ -1,5 +1,4 @@VITE_SUPABASE_URL=https://mpvhvyxuqfcemfnwiaiq.supabase.coVITE_SUPABASE_ANON_KEY=sb_publishable_V-K9coUWXVy8dsZPGKDw-g_d1JzJf8u-# The Diffusion Studio API; http://localhost:3000 for a local oneVITE_API_URL=https://api.diffusion.studiodiff --git a/apps/web/package.json b/apps/web/package.jsonindex 611d0881..b0885ca9 100644--- a/apps/web/package.json+++ b/apps/web/package.json@@ -14,7 +14,7 @@},"dependencies": {"@diffusionstudio/agent-chat": "*",- "@diffusionstudio/api-contract": "^0.2.0",+ "@diffusionstudio/api-contract": "^0.4.0","@diffusionstudio/assets": "*","@diffusionstudio/dapi": "*","@diffusionstudio/encoder": "*",diff --git a/apps/web/src/components/genai/config.ts b/apps/web/src/components/genai/config.tsindex 6d8b28a0..657ab392 100644--- a/apps/web/src/components/genai/config.ts+++ b/apps/web/src/components/genai/config.ts@@ -36,7 +36,7 @@ export const MODEL_MODES: Record<ModelId, PromptMode | "TEXT"> = {"elevenlabs-v3": "VOICE","gemini-3.8-flash-tts": "VOICE","universal-3.5-pro": "TEXT",- "gemini-3.5-flash": "TEXT",+ "qwen3.8-omni-flash": "TEXT",};export type PromptMode = "IMAGE" | "VIDEO" | "VOICE" | "AUDIO";diff --git a/apps/web/src/dapi/handlers/listen.ts b/apps/web/src/dapi/handlers/listen.tsindex 6663ce2c..25bb30c7 100644--- a/apps/web/src/dapi/handlers/listen.ts+++ b/apps/web/src/dapi/handlers/listen.ts@@ -27,7 +27,7 @@ export const listen: ToolHandler<"listen"> = async ({ path, prompt, start, end }try {const media = await uploadFile(windowed ? await audioFile(asset, from, to) : await audioFile(asset));- const job = await runJob({ model: "gemini-3.5-flash", media, prompt }, () => {}, ctx.signal);+ const job = await runJob({ model: "qwen3.8-omni-flash", media, prompt }, () => {}, ctx.signal);if (job.status !== "succeeded") throw new Error(jobFailure(job));const [file] = job.assets;diff --git a/docs/reference/models.md b/docs/reference/models.mdindex ea7feba8..c6e1b5f3 100644--- a/docs/reference/models.md+++ b/docs/reference/models.md@@ -8,7 +8,7 @@ Every model [`generate`](./tools/generate.md) runs, and the fields each takes. A| [Video](#video) | `veo-3.1`, `veo-3.1-fast`, `kling-3-pro`, `kling-o3-pro`, `seedance-2.5`, `wan-3.0`, `minimax-h3`, `flux-3-video`, `grok-imagine-video-1.5` | a clip from a prompt, optionally from a start (and end) frame || [Audio](#audio) | `elevenlabs-music`, `elevenlabs-sfx` | music, sound effects || [Voice](#voice) | `elevenlabs-v3`, `gemini-3.8-flash-tts` | speech |-| [Tools](#tools) | `bria-rmbg-2.0`, `seedvr-2`, `bytedance-upscaler`, `universal-3.5-pro`, `gemini-3.5-flash` | one file in, one file out |+| [Tools](#tools) | `bria-rmbg-2.0`, `seedvr-2`, `bytedance-upscaler`, `universal-3.5-pro`, `qwen3.8-omni-flash` | one file in, one file out || [Local](#local) | `sam-2.1` | runs on this machine, free: an object's mask, tracked through footage |## Image@@ -93,7 +93,7 @@ Each takes one file and returns one.| `seedvr-2` | SeedVR2 — image upscaling | `image`: `{ path }`, required | a PNG of the image upscaled 2× || `bytedance-upscaler` | ByteDance Upscaler — video upscaling | `video`: `{ path }`, required | an MP4 of the video upscaled to 2K at 30 fps || `universal-3.5-pro` | AssemblyAI Universal-3.5 Pro — transcription with speaker labels (Universal-2 for languages it lacks) | `audio`: `{ path }` (audio or video) or `{ scene }` (the scene's mix, see [Scenes](./tools/generate.md#scenes)), required; `languageCode`: e.g. `en`, `en_us`, `de` (detected when left out) | `transcript.json`: `[{ text, speaker?, words: [{ text, start, end }] }]`, times in seconds, speakers labelled `A`, `B`, … when told apart |-| `gemini-3.5-flash` | Gemini 3.5 Flash — describes or answers questions about media | `media`: `{ path }`, required (image, video or audio); `prompt`: what to look for, up to 10,000 characters (a general description when left out) | the analysis, as a file |+| `qwen3.8-omni-flash` | Qwen3.8 Omni Flash — describes or answers questions about audio: speech, speakers, music, sounds | `media`: `{ path }`, required: an audio file (wav, mp3, ogg, flac, aac, m4a or aiff) up to 20 MB; `prompt`: what to listen for, up to 10,000 characters (a general description when left out) | `analysis.md`, the answer as Markdown |## Localdiff --git a/docs/skills/editor.md b/docs/skills/editor.mdindex 60c35b23..3d1713b4 100644--- a/docs/skills/editor.md+++ b/docs/skills/editor.md@@ -52,6 +52,7 @@ How to confirm a change actually produced what you intended. A clean save does n- For tasks that don't need an editing UI, keep the app in the background: from a shell, `diffusion open -b <dir>` launches it that way. Over MCP the app is already running, and `open` only opens the folder.- Only render (export) the result when prompted.- Start with a fresh project.+- Don't open a project in a cloud-synced folder (iCloud Drive, including a synced Desktop or Documents, Dropbox, OneDrive, Google Drive). Sync clients evict media to placeholders, upload half-written renders and leave conflict copies. Use a local folder outside the synced tree.# Docsdiff --git a/packages/dapi/package.json b/packages/dapi/package.jsonindex a90d0f11..c35bb514 100644--- a/packages/dapi/package.json+++ b/packages/dapi/package.json@@ -18,7 +18,7 @@"zod"],"dependencies": {- "@diffusionstudio/api-contract": "^0.3.0",+ "@diffusionstudio/api-contract": "^0.4.0","@diffusionstudio/jsx": "*","zod": "^4.4.3"},