Skip to content

Free, open-source AI video generator for Mac. Run LTX-2.5 and MiniMax H3 (Hailuo) locally on Apple Silicon: text-to-video and image-to-video with sound, lip-sync, character LoRA training, songs from lyrics (YuE2) and image editing (Qwen-Image-Edit). One-click install via Pinokio. Offline: no cloud, no API key, no NVIDIA GPU.

feat(settings): add output codec settings page; default to web-compatible yuv420p

mrbizarro6793ed2f3091
diff --git a/.clinerules b/.clinerules
index 8140aa4..9a84e57 100644
--- a/.clinerules
+++ b/.clinerules
@@ -310,6 +310,7 @@ endpoints are unauthenticated (loopback only).
| `/status` | JSON | Snapshot of `STATE` — running flag, current job, queue, history (top 20), log tail, tier, memory pressure, helper alive flag, comfy PIDs, repo download state. Polled by the UI every ~1s. |
| `/uploads` | JSON | Recent files in `panel_uploads/` for the picker's "click to reuse" strip. |
| `/models` | JSON | Per-repo install completeness — driven by `required_files.json`. |
+| `/settings` | JSON | Returns `{settings, presets, default_preset}`. `settings` is the current `panel_settings.json` content. `presets` is the OUTPUT_PRESETS table (label + blurb + pix_fmt + crf per preset) so the UI doesn't duplicate it. |
| `/file?path=…` | bytes (mp4/png/jpg) | Range-aware video serve so `<video>` tags can seek without re-downloading. |
| `/image?path=…` | bytes (image) | Same idea for image previews. |
| `/sidecar?path=…` | JSON | Per-output sidecar (job params, output stats) if it exists. |
@@ -330,6 +331,7 @@ endpoints are unauthenticated (loopback only).
| `/output/show_all` | (none) | `{unhidden_count: int}` | Clear hidden set. |
| `/upload` | multipart `image=<file>` | `{ok, path}` | Saves to `panel_uploads/<ts>_<safename>`. |
| `/helper/restart` | (none) | `{ok}` | SIGTERM the helper subprocess; the next job auto-respawns it. Useful for picking up site-packages changes. |
+| `/settings` | `output_preset=<key>` (+ optional `output_pix_fmt`, `output_crf` for custom) | `{ok, settings, helper_restarted}` | Update + persist panel settings. If the codec changed, the helper is killed so the next job respawns it with the new env. Validation rejects unknown presets, unknown pix_fmt values, or CRF outside 0–30. |
| `/stop` | (none) | `{ok}` | Cancel the current job (kills helper + mux). Worker advances. |
| `/stop_comfy` | (none) | `{ok, killed}` | SIGTERM any ComfyUI process matching `LTX_COMFY_PATTERN`. |
| `/open_pinokio` | (none) | `{ok}` | macOS-only: focus the Pinokio app. |
@@ -415,6 +417,43 @@ The thresholds are intentional. M-Studio 64 GB is the canonical
"standard" hardware. Anything below 48 GB cannot fit Q8 weights
(transformer-dev.safetensors is 19.18 GiB plus working tensors).
+## 15a. `panel_settings.json` schema
+
+User-controllable preferences. Persisted at the install root, gitignored
+(per-user config, not part of the repo). Read at panel startup; mutated
+via `POST /settings`. The helper subprocess inherits codec env vars from
+this at spawn — when settings change, the panel kills the helper so the
+next job picks up new values.
+
+```jsonc
+{
+ "version": 1,
+ // Metadata so the UI knows which preset to highlight as active.
+ // "custom" means pix_fmt + crf were set manually outside any preset.
+ "output_preset": "standard" | "archival" | "web" | "custom",
+ "output_pix_fmt": "yuv420p" | "yuv444p" | "yuv422p"
+ | "yuv420p10le" | "yuv422p10le" | "yuv444p10le",
+ "output_crf": "0".."30" // string (passed via env vars)
+}
+```
+
+Preset table is the single source of truth, defined in `OUTPUT_PRESETS`
+in `mlx_ltx_panel.py`:
+
+| key | pix_fmt | crf | Approx 5s @ 1280×704 | Use case |
+|---|---|---|---|---|
+| `standard` ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless, plays everywhere (X / IG / Discord) |
+| `archival` | yuv444p | 0 | ~50 MB | Mathematically lossless, post-processing master |
+| `web` | yuv420p | 23 | ~3 MB | Smallest, mobile / embedding |
+
+`+faststart` is always applied (codec patch) regardless of preset, so
+gallery thumbnails always render the first frame instantly.
+
+When adding a new preset: edit `OUTPUT_PRESETS` in `mlx_ltx_panel.py`
+and the `order` array in the settings modal JS (`openSettingsModal`).
+The blurb shown in the UI is read from the server at modal-open time,
+so server is the source of truth — no JS strings to keep in sync.
+
## 16. `required_files.json` schema
Single source of truth for "what counts as installed." Read by:
diff --git a/.cursorrules b/.cursorrules
index 8140aa4..9a84e57 100644
--- a/.cursorrules
+++ b/.cursorrules
@@ -310,6 +310,7 @@ endpoints are unauthenticated (loopback only).
| `/status` | JSON | Snapshot of `STATE` — running flag, current job, queue, history (top 20), log tail, tier, memory pressure, helper alive flag, comfy PIDs, repo download state. Polled by the UI every ~1s. |
| `/uploads` | JSON | Recent files in `panel_uploads/` for the picker's "click to reuse" strip. |
| `/models` | JSON | Per-repo install completeness — driven by `required_files.json`. |
+| `/settings` | JSON | Returns `{settings, presets, default_preset}`. `settings` is the current `panel_settings.json` content. `presets` is the OUTPUT_PRESETS table (label + blurb + pix_fmt + crf per preset) so the UI doesn't duplicate it. |
| `/file?path=…` | bytes (mp4/png/jpg) | Range-aware video serve so `<video>` tags can seek without re-downloading. |
| `/image?path=…` | bytes (image) | Same idea for image previews. |
| `/sidecar?path=…` | JSON | Per-output sidecar (job params, output stats) if it exists. |
@@ -330,6 +331,7 @@ endpoints are unauthenticated (loopback only).
| `/output/show_all` | (none) | `{unhidden_count: int}` | Clear hidden set. |
| `/upload` | multipart `image=<file>` | `{ok, path}` | Saves to `panel_uploads/<ts>_<safename>`. |
| `/helper/restart` | (none) | `{ok}` | SIGTERM the helper subprocess; the next job auto-respawns it. Useful for picking up site-packages changes. |
+| `/settings` | `output_preset=<key>` (+ optional `output_pix_fmt`, `output_crf` for custom) | `{ok, settings, helper_restarted}` | Update + persist panel settings. If the codec changed, the helper is killed so the next job respawns it with the new env. Validation rejects unknown presets, unknown pix_fmt values, or CRF outside 0–30. |
| `/stop` | (none) | `{ok}` | Cancel the current job (kills helper + mux). Worker advances. |
| `/stop_comfy` | (none) | `{ok, killed}` | SIGTERM any ComfyUI process matching `LTX_COMFY_PATTERN`. |
| `/open_pinokio` | (none) | `{ok}` | macOS-only: focus the Pinokio app. |
@@ -415,6 +417,43 @@ The thresholds are intentional. M-Studio 64 GB is the canonical
"standard" hardware. Anything below 48 GB cannot fit Q8 weights
(transformer-dev.safetensors is 19.18 GiB plus working tensors).
+## 15a. `panel_settings.json` schema
+
+User-controllable preferences. Persisted at the install root, gitignored
+(per-user config, not part of the repo). Read at panel startup; mutated
+via `POST /settings`. The helper subprocess inherits codec env vars from
+this at spawn — when settings change, the panel kills the helper so the
+next job picks up new values.
+
+```jsonc
+{
+ "version": 1,
+ // Metadata so the UI knows which preset to highlight as active.
+ // "custom" means pix_fmt + crf were set manually outside any preset.
+ "output_preset": "standard" | "archival" | "web" | "custom",
+ "output_pix_fmt": "yuv420p" | "yuv444p" | "yuv422p"
+ | "yuv420p10le" | "yuv422p10le" | "yuv444p10le",
+ "output_crf": "0".."30" // string (passed via env vars)
+}
+```
+
+Preset table is the single source of truth, defined in `OUTPUT_PRESETS`
+in `mlx_ltx_panel.py`:
+
+| key | pix_fmt | crf | Approx 5s @ 1280×704 | Use case |
+|---|---|---|---|---|
+| `standard` ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless, plays everywhere (X / IG / Discord) |
+| `archival` | yuv444p | 0 | ~50 MB | Mathematically lossless, post-processing master |
+| `web` | yuv420p | 23 | ~3 MB | Smallest, mobile / embedding |
+
+`+faststart` is always applied (codec patch) regardless of preset, so
+gallery thumbnails always render the first frame instantly.
+
+When adding a new preset: edit `OUTPUT_PRESETS` in `mlx_ltx_panel.py`
+and the `order` array in the settings modal JS (`openSettingsModal`).
+The blurb shown in the UI is read from the server at modal-open time,
+so server is the source of truth — no JS strings to keep in sync.
+
## 16. `required_files.json` schema
Single source of truth for "what counts as installed." Read by:
diff --git a/.gitignore b/.gitignore
index d79c98e..666fe12 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,6 +3,7 @@
# =============================================================================
panel_queue.json
panel_hidden.json
+panel_settings.json
*.log
*.tmp
diff --git a/.windsurfrules b/.windsurfrules
index 8140aa4..9a84e57 100644
--- a/.windsurfrules
+++ b/.windsurfrules
@@ -310,6 +310,7 @@ endpoints are unauthenticated (loopback only).
| `/status` | JSON | Snapshot of `STATE` — running flag, current job, queue, history (top 20), log tail, tier, memory pressure, helper alive flag, comfy PIDs, repo download state. Polled by the UI every ~1s. |
| `/uploads` | JSON | Recent files in `panel_uploads/` for the picker's "click to reuse" strip. |
| `/models` | JSON | Per-repo install completeness — driven by `required_files.json`. |
+| `/settings` | JSON | Returns `{settings, presets, default_preset}`. `settings` is the current `panel_settings.json` content. `presets` is the OUTPUT_PRESETS table (label + blurb + pix_fmt + crf per preset) so the UI doesn't duplicate it. |
| `/file?path=…` | bytes (mp4/png/jpg) | Range-aware video serve so `<video>` tags can seek without re-downloading. |
| `/image?path=…` | bytes (image) | Same idea for image previews. |
| `/sidecar?path=…` | JSON | Per-output sidecar (job params, output stats) if it exists. |
@@ -330,6 +331,7 @@ endpoints are unauthenticated (loopback only).
| `/output/show_all` | (none) | `{unhidden_count: int}` | Clear hidden set. |
| `/upload` | multipart `image=<file>` | `{ok, path}` | Saves to `panel_uploads/<ts>_<safename>`. |
| `/helper/restart` | (none) | `{ok}` | SIGTERM the helper subprocess; the next job auto-respawns it. Useful for picking up site-packages changes. |
+| `/settings` | `output_preset=<key>` (+ optional `output_pix_fmt`, `output_crf` for custom) | `{ok, settings, helper_restarted}` | Update + persist panel settings. If the codec changed, the helper is killed so the next job respawns it with the new env. Validation rejects unknown presets, unknown pix_fmt values, or CRF outside 0–30. |
| `/stop` | (none) | `{ok}` | Cancel the current job (kills helper + mux). Worker advances. |
| `/stop_comfy` | (none) | `{ok, killed}` | SIGTERM any ComfyUI process matching `LTX_COMFY_PATTERN`. |
| `/open_pinokio` | (none) | `{ok}` | macOS-only: focus the Pinokio app. |
@@ -415,6 +417,43 @@ The thresholds are intentional. M-Studio 64 GB is the canonical
"standard" hardware. Anything below 48 GB cannot fit Q8 weights
(transformer-dev.safetensors is 19.18 GiB plus working tensors).
+## 15a. `panel_settings.json` schema
+
+User-controllable preferences. Persisted at the install root, gitignored
+(per-user config, not part of the repo). Read at panel startup; mutated
+via `POST /settings`. The helper subprocess inherits codec env vars from
+this at spawn — when settings change, the panel kills the helper so the
+next job picks up new values.
+
+```jsonc
+{
+ "version": 1,
+ // Metadata so the UI knows which preset to highlight as active.
+ // "custom" means pix_fmt + crf were set manually outside any preset.
+ "output_preset": "standard" | "archival" | "web" | "custom",
+ "output_pix_fmt": "yuv420p" | "yuv444p" | "yuv422p"
+ | "yuv420p10le" | "yuv422p10le" | "yuv444p10le",
+ "output_crf": "0".."30" // string (passed via env vars)
+}
+```
+
+Preset table is the single source of truth, defined in `OUTPUT_PRESETS`
+in `mlx_ltx_panel.py`:
+
+| key | pix_fmt | crf | Approx 5s @ 1280×704 | Use case |
+|---|---|---|---|---|
+| `standard` ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless, plays everywhere (X / IG / Discord) |
+| `archival` | yuv444p | 0 | ~50 MB | Mathematically lossless, post-processing master |
+| `web` | yuv420p | 23 | ~3 MB | Smallest, mobile / embedding |
+
+`+faststart` is always applied (codec patch) regardless of preset, so
+gallery thumbnails always render the first frame instantly.
+
+When adding a new preset: edit `OUTPUT_PRESETS` in `mlx_ltx_panel.py`
+and the `order` array in the settings modal JS (`openSettingsModal`).
+The blurb shown in the UI is read from the server at modal-open time,
+so server is the source of truth — no JS strings to keep in sync.
+
## 16. `required_files.json` schema
Single source of truth for "what counts as installed." Read by:
diff --git a/AGENTS.md b/AGENTS.md
index 8140aa4..9a84e57 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -310,6 +310,7 @@ endpoints are unauthenticated (loopback only).
| `/status` | JSON | Snapshot of `STATE` — running flag, current job, queue, history (top 20), log tail, tier, memory pressure, helper alive flag, comfy PIDs, repo download state. Polled by the UI every ~1s. |
| `/uploads` | JSON | Recent files in `panel_uploads/` for the picker's "click to reuse" strip. |
| `/models` | JSON | Per-repo install completeness — driven by `required_files.json`. |
+| `/settings` | JSON | Returns `{settings, presets, default_preset}`. `settings` is the current `panel_settings.json` content. `presets` is the OUTPUT_PRESETS table (label + blurb + pix_fmt + crf per preset) so the UI doesn't duplicate it. |
| `/file?path=…` | bytes (mp4/png/jpg) | Range-aware video serve so `<video>` tags can seek without re-downloading. |
| `/image?path=…` | bytes (image) | Same idea for image previews. |
| `/sidecar?path=…` | JSON | Per-output sidecar (job params, output stats) if it exists. |
@@ -330,6 +331,7 @@ endpoints are unauthenticated (loopback only).
| `/output/show_all` | (none) | `{unhidden_count: int}` | Clear hidden set. |
| `/upload` | multipart `image=<file>` | `{ok, path}` | Saves to `panel_uploads/<ts>_<safename>`. |
| `/helper/restart` | (none) | `{ok}` | SIGTERM the helper subprocess; the next job auto-respawns it. Useful for picking up site-packages changes. |
+| `/settings` | `output_preset=<key>` (+ optional `output_pix_fmt`, `output_crf` for custom) | `{ok, settings, helper_restarted}` | Update + persist panel settings. If the codec changed, the helper is killed so the next job respawns it with the new env. Validation rejects unknown presets, unknown pix_fmt values, or CRF outside 0–30. |
| `/stop` | (none) | `{ok}` | Cancel the current job (kills helper + mux). Worker advances. |
| `/stop_comfy` | (none) | `{ok, killed}` | SIGTERM any ComfyUI process matching `LTX_COMFY_PATTERN`. |
| `/open_pinokio` | (none) | `{ok}` | macOS-only: focus the Pinokio app. |
@@ -415,6 +417,43 @@ The thresholds are intentional. M-Studio 64 GB is the canonical
"standard" hardware. Anything below 48 GB cannot fit Q8 weights
(transformer-dev.safetensors is 19.18 GiB plus working tensors).
+## 15a. `panel_settings.json` schema
+
+User-controllable preferences. Persisted at the install root, gitignored
+(per-user config, not part of the repo). Read at panel startup; mutated
+via `POST /settings`. The helper subprocess inherits codec env vars from
+this at spawn — when settings change, the panel kills the helper so the
+next job picks up new values.
+
+```jsonc
+{
+ "version": 1,
+ // Metadata so the UI knows which preset to highlight as active.
+ // "custom" means pix_fmt + crf were set manually outside any preset.
+ "output_preset": "standard" | "archival" | "web" | "custom",
+ "output_pix_fmt": "yuv420p" | "yuv444p" | "yuv422p"
+ | "yuv420p10le" | "yuv422p10le" | "yuv444p10le",
+ "output_crf": "0".."30" // string (passed via env vars)
+}
+```
+
+Preset table is the single source of truth, defined in `OUTPUT_PRESETS`
+in `mlx_ltx_panel.py`:
+
+| key | pix_fmt | crf | Approx 5s @ 1280×704 | Use case |
+|---|---|---|---|---|
+| `standard` ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless, plays everywhere (X / IG / Discord) |
+| `archival` | yuv444p | 0 | ~50 MB | Mathematically lossless, post-processing master |
+| `web` | yuv420p | 23 | ~3 MB | Smallest, mobile / embedding |
+
+`+faststart` is always applied (codec patch) regardless of preset, so
+gallery thumbnails always render the first frame instantly.
+
+When adding a new preset: edit `OUTPUT_PRESETS` in `mlx_ltx_panel.py`
+and the `order` array in the settings modal JS (`openSettingsModal`).
+The blurb shown in the UI is read from the server at modal-open time,
+so server is the source of truth — no JS strings to keep in sync.
+
## 16. `required_files.json` schema
Single source of truth for "what counts as installed." Read by:
diff --git a/CLAUDE.md b/CLAUDE.md
index 8140aa4..9a84e57 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -310,6 +310,7 @@ endpoints are unauthenticated (loopback only).
| `/status` | JSON | Snapshot of `STATE` — running flag, current job, queue, history (top 20), log tail, tier, memory pressure, helper alive flag, comfy PIDs, repo download state. Polled by the UI every ~1s. |
| `/uploads` | JSON | Recent files in `panel_uploads/` for the picker's "click to reuse" strip. |
| `/models` | JSON | Per-repo install completeness — driven by `required_files.json`. |
+| `/settings` | JSON | Returns `{settings, presets, default_preset}`. `settings` is the current `panel_settings.json` content. `presets` is the OUTPUT_PRESETS table (label + blurb + pix_fmt + crf per preset) so the UI doesn't duplicate it. |
| `/file?path=…` | bytes (mp4/png/jpg) | Range-aware video serve so `<video>` tags can seek without re-downloading. |
| `/image?path=…` | bytes (image) | Same idea for image previews. |
| `/sidecar?path=…` | JSON | Per-output sidecar (job params, output stats) if it exists. |
@@ -330,6 +331,7 @@ endpoints are unauthenticated (loopback only).
| `/output/show_all` | (none) | `{unhidden_count: int}` | Clear hidden set. |
| `/upload` | multipart `image=<file>` | `{ok, path}` | Saves to `panel_uploads/<ts>_<safename>`. |
| `/helper/restart` | (none) | `{ok}` | SIGTERM the helper subprocess; the next job auto-respawns it. Useful for picking up site-packages changes. |
+| `/settings` | `output_preset=<key>` (+ optional `output_pix_fmt`, `output_crf` for custom) | `{ok, settings, helper_restarted}` | Update + persist panel settings. If the codec changed, the helper is killed so the next job respawns it with the new env. Validation rejects unknown presets, unknown pix_fmt values, or CRF outside 0–30. |
| `/stop` | (none) | `{ok}` | Cancel the current job (kills helper + mux). Worker advances. |
| `/stop_comfy` | (none) | `{ok, killed}` | SIGTERM any ComfyUI process matching `LTX_COMFY_PATTERN`. |
| `/open_pinokio` | (none) | `{ok}` | macOS-only: focus the Pinokio app. |
@@ -415,6 +417,43 @@ The thresholds are intentional. M-Studio 64 GB is the canonical
"standard" hardware. Anything below 48 GB cannot fit Q8 weights
(transformer-dev.safetensors is 19.18 GiB plus working tensors).
+## 15a. `panel_settings.json` schema
+
+User-controllable preferences. Persisted at the install root, gitignored
+(per-user config, not part of the repo). Read at panel startup; mutated
+via `POST /settings`. The helper subprocess inherits codec env vars from
+this at spawn — when settings change, the panel kills the helper so the
+next job picks up new values.
+
+```jsonc
+{
+ "version": 1,
+ // Metadata so the UI knows which preset to highlight as active.
+ // "custom" means pix_fmt + crf were set manually outside any preset.
+ "output_preset": "standard" | "archival" | "web" | "custom",
+ "output_pix_fmt": "yuv420p" | "yuv444p" | "yuv422p"
+ | "yuv420p10le" | "yuv422p10le" | "yuv444p10le",
+ "output_crf": "0".."30" // string (passed via env vars)
+}
+```
+
+Preset table is the single source of truth, defined in `OUTPUT_PRESETS`
+in `mlx_ltx_panel.py`:
+
+| key | pix_fmt | crf | Approx 5s @ 1280×704 | Use case |
+|---|---|---|---|---|
+| `standard` ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless, plays everywhere (X / IG / Discord) |
+| `archival` | yuv444p | 0 | ~50 MB | Mathematically lossless, post-processing master |
+| `web` | yuv420p | 23 | ~3 MB | Smallest, mobile / embedding |
+
+`+faststart` is always applied (codec patch) regardless of preset, so
+gallery thumbnails always render the first frame instantly.
+
+When adding a new preset: edit `OUTPUT_PRESETS` in `mlx_ltx_panel.py`
+and the `order` array in the settings modal JS (`openSettingsModal`).
+The blurb shown in the UI is read from the server at modal-open time,
+so server is the source of truth — no JS strings to keep in sync.
+
## 16. `required_files.json` schema
Single source of truth for "what counts as installed." Read by:
diff --git a/GEMINI.md b/GEMINI.md
index 8140aa4..9a84e57 100644
--- a/GEMINI.md
+++ b/GEMINI.md
@@ -310,6 +310,7 @@ endpoints are unauthenticated (loopback only).
| `/status` | JSON | Snapshot of `STATE` — running flag, current job, queue, history (top 20), log tail, tier, memory pressure, helper alive flag, comfy PIDs, repo download state. Polled by the UI every ~1s. |
| `/uploads` | JSON | Recent files in `panel_uploads/` for the picker's "click to reuse" strip. |
| `/models` | JSON | Per-repo install completeness — driven by `required_files.json`. |
+| `/settings` | JSON | Returns `{settings, presets, default_preset}`. `settings` is the current `panel_settings.json` content. `presets` is the OUTPUT_PRESETS table (label + blurb + pix_fmt + crf per preset) so the UI doesn't duplicate it. |
| `/file?path=…` | bytes (mp4/png/jpg) | Range-aware video serve so `<video>` tags can seek without re-downloading. |
| `/image?path=…` | bytes (image) | Same idea for image previews. |
| `/sidecar?path=…` | JSON | Per-output sidecar (job params, output stats) if it exists. |
@@ -330,6 +331,7 @@ endpoints are unauthenticated (loopback only).
| `/output/show_all` | (none) | `{unhidden_count: int}` | Clear hidden set. |
| `/upload` | multipart `image=<file>` | `{ok, path}` | Saves to `panel_uploads/<ts>_<safename>`. |
| `/helper/restart` | (none) | `{ok}` | SIGTERM the helper subprocess; the next job auto-respawns it. Useful for picking up site-packages changes. |
+| `/settings` | `output_preset=<key>` (+ optional `output_pix_fmt`, `output_crf` for custom) | `{ok, settings, helper_restarted}` | Update + persist panel settings. If the codec changed, the helper is killed so the next job respawns it with the new env. Validation rejects unknown presets, unknown pix_fmt values, or CRF outside 0–30. |
| `/stop` | (none) | `{ok}` | Cancel the current job (kills helper + mux). Worker advances. |
| `/stop_comfy` | (none) | `{ok, killed}` | SIGTERM any ComfyUI process matching `LTX_COMFY_PATTERN`. |
| `/open_pinokio` | (none) | `{ok}` | macOS-only: focus the Pinokio app. |
@@ -415,6 +417,43 @@ The thresholds are intentional. M-Studio 64 GB is the canonical
"standard" hardware. Anything below 48 GB cannot fit Q8 weights
(transformer-dev.safetensors is 19.18 GiB plus working tensors).
+## 15a. `panel_settings.json` schema
+
+User-controllable preferences. Persisted at the install root, gitignored
+(per-user config, not part of the repo). Read at panel startup; mutated
+via `POST /settings`. The helper subprocess inherits codec env vars from
+this at spawn — when settings change, the panel kills the helper so the
+next job picks up new values.
+
+```jsonc
+{
+ "version": 1,
+ // Metadata so the UI knows which preset to highlight as active.
+ // "custom" means pix_fmt + crf were set manually outside any preset.
+ "output_preset": "standard" | "archival" | "web" | "custom",
+ "output_pix_fmt": "yuv420p" | "yuv444p" | "yuv422p"
+ | "yuv420p10le" | "yuv422p10le" | "yuv444p10le",
+ "output_crf": "0".."30" // string (passed via env vars)
+}
+```
+
+Preset table is the single source of truth, defined in `OUTPUT_PRESETS`
+in `mlx_ltx_panel.py`:
+
+| key | pix_fmt | crf | Approx 5s @ 1280×704 | Use case |
+|---|---|---|---|---|
+| `standard` ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless, plays everywhere (X / IG / Discord) |
+| `archival` | yuv444p | 0 | ~50 MB | Mathematically lossless, post-processing master |
+| `web` | yuv420p | 23 | ~3 MB | Smallest, mobile / embedding |
+
+`+faststart` is always applied (codec patch) regardless of preset, so
+gallery thumbnails always render the first frame instantly.
+
+When adding a new preset: edit `OUTPUT_PRESETS` in `mlx_ltx_panel.py`
+and the `order` array in the settings modal JS (`openSettingsModal`).
+The blurb shown in the UI is read from the server at modal-open time,
+so server is the source of truth — no JS strings to keep in sync.
+
## 16. `required_files.json` schema
Single source of truth for "what counts as installed." Read by:
diff --git a/QWEN.md b/QWEN.md
index 8140aa4..9a84e57 100644
--- a/QWEN.md
+++ b/QWEN.md
@@ -310,6 +310,7 @@ endpoints are unauthenticated (loopback only).
| `/status` | JSON | Snapshot of `STATE` — running flag, current job, queue, history (top 20), log tail, tier, memory pressure, helper alive flag, comfy PIDs, repo download state. Polled by the UI every ~1s. |
| `/uploads` | JSON | Recent files in `panel_uploads/` for the picker's "click to reuse" strip. |
| `/models` | JSON | Per-repo install completeness — driven by `required_files.json`. |
+| `/settings` | JSON | Returns `{settings, presets, default_preset}`. `settings` is the current `panel_settings.json` content. `presets` is the OUTPUT_PRESETS table (label + blurb + pix_fmt + crf per preset) so the UI doesn't duplicate it. |
| `/file?path=…` | bytes (mp4/png/jpg) | Range-aware video serve so `<video>` tags can seek without re-downloading. |
| `/image?path=…` | bytes (image) | Same idea for image previews. |
| `/sidecar?path=…` | JSON | Per-output sidecar (job params, output stats) if it exists. |
@@ -330,6 +331,7 @@ endpoints are unauthenticated (loopback only).
| `/output/show_all` | (none) | `{unhidden_count: int}` | Clear hidden set. |
| `/upload` | multipart `image=<file>` | `{ok, path}` | Saves to `panel_uploads/<ts>_<safename>`. |
| `/helper/restart` | (none) | `{ok}` | SIGTERM the helper subprocess; the next job auto-respawns it. Useful for picking up site-packages changes. |
+| `/settings` | `output_preset=<key>` (+ optional `output_pix_fmt`, `output_crf` for custom) | `{ok, settings, helper_restarted}` | Update + persist panel settings. If the codec changed, the helper is killed so the next job respawns it with the new env. Validation rejects unknown presets, unknown pix_fmt values, or CRF outside 0–30. |
| `/stop` | (none) | `{ok}` | Cancel the current job (kills helper + mux). Worker advances. |
| `/stop_comfy` | (none) | `{ok, killed}` | SIGTERM any ComfyUI process matching `LTX_COMFY_PATTERN`. |
| `/open_pinokio` | (none) | `{ok}` | macOS-only: focus the Pinokio app. |
@@ -415,6 +417,43 @@ The thresholds are intentional. M-Studio 64 GB is the canonical
"standard" hardware. Anything below 48 GB cannot fit Q8 weights
(transformer-dev.safetensors is 19.18 GiB plus working tensors).
+## 15a. `panel_settings.json` schema
+
+User-controllable preferences. Persisted at the install root, gitignored
+(per-user config, not part of the repo). Read at panel startup; mutated
+via `POST /settings`. The helper subprocess inherits codec env vars from
+this at spawn — when settings change, the panel kills the helper so the
+next job picks up new values.
+
+```jsonc
+{
+ "version": 1,
+ // Metadata so the UI knows which preset to highlight as active.
+ // "custom" means pix_fmt + crf were set manually outside any preset.
+ "output_preset": "standard" | "archival" | "web" | "custom",
+ "output_pix_fmt": "yuv420p" | "yuv444p" | "yuv422p"
+ | "yuv420p10le" | "yuv422p10le" | "yuv444p10le",
+ "output_crf": "0".."30" // string (passed via env vars)
+}
+```
+
+Preset table is the single source of truth, defined in `OUTPUT_PRESETS`
+in `mlx_ltx_panel.py`:
+
+| key | pix_fmt | crf | Approx 5s @ 1280×704 | Use case |
+|---|---|---|---|---|
+| `standard` ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless, plays everywhere (X / IG / Discord) |
+| `archival` | yuv444p | 0 | ~50 MB | Mathematically lossless, post-processing master |
+| `web` | yuv420p | 23 | ~3 MB | Smallest, mobile / embedding |
+
+`+faststart` is always applied (codec patch) regardless of preset, so
+gallery thumbnails always render the first frame instantly.
+
+When adding a new preset: edit `OUTPUT_PRESETS` in `mlx_ltx_panel.py`
+and the `order` array in the settings modal JS (`openSettingsModal`).
+The blurb shown in the UI is read from the server at modal-open time,
+so server is the source of truth — no JS strings to keep in sync.
+
## 16. `required_files.json` schema
Single source of truth for "what counts as installed." Read by:
diff --git a/README.md b/README.md
index 47377ac..75c9d7a 100644
--- a/README.md
+++ b/README.md
@@ -194,15 +194,23 @@ Enhance — it appends an audio constraint to the prompt at submit time.
## Output format
-**Lossless H.264 by default** — `yuv444p`, `crf 0`. Your archive is
-the highest fidelity the renderer can produce. Web and social
-platforms will re-encode anyway. Override via env vars
-(`LTX_OUTPUT_PIX_FMT`, `LTX_OUTPUT_CRF`) if you want `yuv420p`
-directly.
-
-**`+faststart`** is on. The `moov` atom sits at the front of the
-file so gallery thumbnails render the first frame instantly without
-downloading the full clip.
+Pick a preset from the **⚙ Settings** modal in the panel header. Three
+built-ins, plus a Custom mode for advanced overrides:
+
+| Preset | pix_fmt | CRF | ~Size (5s @ 1280×704) | Use case |
+|---|---|---|---|---|
+| **Standard** ⭐ default | yuv420p | 18 | ~7 MB | Visually lossless to most viewers. Plays everywhere — X, Instagram, Discord. The default for new installs. |
+| **Archival** | yuv444p | 0 | ~50 MB | Mathematically lossless. Use when you'll re-encode in post or need a master. |
+| **Web** | yuv420p | 23 | ~3 MB | Smallest files. For mobile, embedding, or quick previews. |
+| **Custom** | choose | 0–30 | varies | 10-bit HDR, format-specific delivery, etc. |
+
+Settings persist to `panel_settings.json` and apply to every new
+render. The helper subprocess restarts automatically when you change
+codec settings, so the change takes effect on the next render.
+
+**`+faststart`** is always on, regardless of preset. The `moov` atom
+sits at the front of the file so gallery thumbnails render the first
+frame instantly without downloading the full clip.
For social uploads (X especially rejects `yuv444p`), re-encode with:
diff --git a/mlx_ltx_panel.py b/mlx_ltx_panel.py
index 2059b3b..472e847 100644
--- a/mlx_ltx_panel.py
+++ b/mlx_ltx_panel.py
@@ -110,6 +110,187 @@ _REQUIRED = _load_required_files()
_MIN_FILE_BYTES = int(_REQUIRED.get("min_size_bytes", 1024))
+# ---- panel settings: user-controllable preferences -------------------------
+# Persisted to panel_settings.json so the user's choice survives restarts.
+# Read at startup, exposed via /settings GET, mutated via /settings POST.
+# Currently covers output codec (the cause of the original "X rejects my
+# upload" complaint and the "videos are huge" complaint — lossless yuv444p
+# crf 0 is great for archives but wrong as a default for social workflows).
+#
+# Schema:
+# {
+# "version": 1,
+# "output_preset": "standard" | "archival" | "web" | "custom",
+# "output_pix_fmt": "yuv420p" | "yuv444p" | <other valid ffmpeg pix_fmt>,
+# "output_crf": "0" .. "30" (string, since we pass it via env vars)
+# }
+#
+# `output_preset` is metadata so the UI can show which preset is active even
+# after a reload. The actual ffmpeg behavior is driven by `output_pix_fmt`
+# and `output_crf` which get exported into the helper subprocess's env.
+
+# Preset table — single source of truth for the UI and the install default.
+# Sizes are rough estimates for a 5s 1280x704 H.264 clip (yuv420p crf 18 is
+# the long-standing "visually lossless" default for web video).
+OUTPUT_PRESETS: dict[str, dict[str, str]] = {
+ "archival": {
+ "pix_fmt": "yuv444p",
+ "crf": "0",
+ "label": "Archival (lossless)",
+ "blurb": "Mathematically lossless. ~50 MB per 5s clip. Use for "
+ "post-processing or archive masters.",
+ },
+ "standard": {
+ "pix_fmt": "yuv420p",
+ "crf": "18",
+ "label": "Standard",
+ "blurb": "Visually lossless to ~95% of viewers. ~7 MB per 5s clip. "
+ "Plays everywhere including X / Instagram / Discord. The "
+ "default for new installs.",
+ },
+ "web": {
+ "pix_fmt": "yuv420p",
+ "crf": "23",
+ "label": "Web / social",
+ "blurb": "Smallest files. ~3 MB per 5s clip. For embedding, mobile, "
+ "or quick previews where bandwidth matters more than peak "
+ "fidelity.",
+ },
+}
+
+# Default preset for fresh installs. Switched from "archival" to "standard"
+# after multiple users reported X upload failures (X rejects yuv444p) and
+# disk-fill complaints about ~50 MB clips. Existing installs keep whatever
+# is in their panel_settings.json; new installs and panels with no settings
+# file yet get this default.
+DEFAULT_OUTPUT_PRESET = "standard"
+
+SETTINGS_FILE = ROOT / "panel_settings.json"
+_SETTINGS_LOCK = threading.Lock()
+
+
+def _settings_defaults() -> dict:
+ preset = OUTPUT_PRESETS[DEFAULT_OUTPUT_PRESET]
+ return {
+ "version": 1,
+ "output_preset": DEFAULT_OUTPUT_PRESET,
+ "output_pix_fmt": preset["pix_fmt"],
+ "output_crf": preset["crf"],
+ }
+
+
+def _load_settings() -> dict:
+ """Read panel_settings.json. Missing file → return + write the default
+ so first-run users get the sensible Standard preset. Corrupt file →
+ fall back to defaults but DON'T overwrite (preserves the bad file for
+ forensic inspection if it was edited by hand)."""
+ if not SETTINGS_FILE.exists():
+ defaults = _settings_defaults()
+ try:
+ SETTINGS_FILE.write_text(json.dumps(defaults, indent=2))
+ except Exception as exc:
+ sys.stderr.write(f"WARN: could not write {SETTINGS_FILE} ({exc})\n")
+ return defaults
+ try:
+ with SETTINGS_FILE.open("r") as fh:
+ data = json.load(fh)
+ # Backfill missing keys against defaults so older settings files
+ # don't trip over fields added in later versions.
+ defaults = _settings_defaults()
+ for k, v in defaults.items():
+ data.setdefault(k, v)
+ return data
+ except Exception as exc:
+ sys.stderr.write(f"WARN: panel_settings.json unreadable ({exc}); "
+ f"using defaults until manually fixed\n")
+ return _settings_defaults()
+
+
+def _save_settings(settings: dict) -> None:
+ """Atomic write so a Pinokio kill mid-write can't leave a half-file."""
+ import tempfile, os as _os
+ fd, tmp = tempfile.mkstemp(prefix="panel_settings.", dir=str(ROOT))
+ try:
+ with _os.fdopen(fd, "w") as fh:
+ fh.write(json.dumps(settings, indent=2))
+ fh.flush()
+ _os.fsync(fh.fileno())
+ _os.replace(tmp, SETTINGS_FILE)
+ except Exception:
+ try: _os.unlink(tmp)
+ except OSError: pass
+ raise
+
+
+def _validate_settings_patch(patch: dict) -> tuple[dict, str | None]:
+ """Validate user-submitted partial settings, return (clean, error_or_None).
+ Whitelist what we accept — never trust the form payload to be safe to
+ pass to ffmpeg / shell."""
+ out: dict = {}
+
+ if "output_preset" in patch:
+ preset = str(patch["output_preset"]).strip().lower()
+ if preset not in OUTPUT_PRESETS and preset != "custom":
+ return {}, f"unknown output_preset: {preset}"
+ out["output_preset"] = preset
+ if preset != "custom":
+ # Preset overrides any pix_fmt/crf in the same patch — picking
+ # "Standard" should always give Standard's values.
+ out["output_pix_fmt"] = OUTPUT_PRESETS[preset]["pix_fmt"]
+ out["output_crf"] = OUTPUT_PRESETS[preset]["crf"]
+ return out, None
+
+ # Custom path — validate pix_fmt + crf manually.
+ if "output_pix_fmt" in patch:
+ pf = str(patch["output_pix_fmt"]).strip().lower()
+ # ffmpeg has many pix_fmts; whitelist the ones LTX 2.3 actually
+ # produces correctly. Anything else is a footgun (color shifts,
+ # encoder errors).
+ ALLOWED_PIX_FMTS = {"yuv420p", "yuv422p", "yuv444p", "yuv420p10le",
+ "yuv422p10le", "yuv444p10le"}
+ if pf not in ALLOWED_PIX_FMTS:
+ return {}, (f"output_pix_fmt must be one of: "
+ f"{sorted(ALLOWED_PIX_FMTS)}")
+ out["output_pix_fmt"] = pf
+
+ if "output_crf" in patch:
+ try:
+ crf_i = int(str(patch["output_crf"]))
+ except (TypeError, ValueError):
+ return {}, "output_crf must be an integer 0-30"
+ if not 0 <= crf_i <= 30:
+ return {}, "output_crf must be between 0 (lossless) and 30 (low)"
+ out["output_crf"] = str(crf_i)
+
+ return out, None
+
+
+_SETTINGS = _load_settings()
+
+
+def get_settings() -> dict:
+ with _SETTINGS_LOCK:
+ return dict(_SETTINGS)
+
+
+def update_settings(patch: dict) -> tuple[dict, str | None]:
+ """Apply a validated patch + persist + return (current, error_or_None).
+ Caller is responsible for triggering a helper restart if codec settings
+ changed (the ffmpeg call inside the helper reads env vars at job time,
+ and the helper inherits env at spawn — so a running helper will keep
+ using whatever was active when it spawned)."""
+ clean, err = _validate_settings_patch(patch)
+ if err:
+ return get_settings(), err
+ with _SETTINGS_LOCK:
+ _SETTINGS.update(clean)
+ try:
+ _save_settings(_SETTINGS)
+ except Exception as exc:
+ return get_settings(), f"could not persist settings: {exc}"
+ return dict(_SETTINGS), None
+
+
def _repos() -> list[dict]:
"""All repo entries from required_files.json (list, in order)."""
return list(_REQUIRED.get("repos", []))
@@ -970,6 +1151,14 @@ class WarmHelper:
env["LTX_GEMMA"] = str(GEMMA)
env["LTX_IDLE_TIMEOUT"] = str(HELPER_IDLE_TIMEOUT)
env["LTX_LOW_MEMORY"] = HELPER_LOW_MEMORY
+ # Output codec env vars sourced from panel settings. The patched
+ # ffmpeg call inside ltx_core_mlx reads these at job time, so
+ # they need to be in the helper's env at spawn. When the user
+ # changes settings via /settings POST we kill the helper; the
+ # next job spawns it fresh and picks up the new values here.
+ _s = get_settings()
+ env["LTX_OUTPUT_PIX_FMT"] = _s.get("output_pix_fmt", "yuv420p")
+ env["LTX_OUTPUT_CRF"] = _s.get("output_crf", "18")
push(f"Spawning warm helper (low_memory={HELPER_LOW_MEMORY}, idle_timeout={HELPER_IDLE_TIMEOUT}s)")
self.proc = subprocess.Popen(
[str(HELPER_PYTHON), str(HELPER_SCRIPT)],
@@ -1825,6 +2014,16 @@ class Handler(BaseHTTPRequestHandler):
)
self._json(payload)
return
+ if parsed.path == "/settings":
+ # Return current panel settings + the preset table so the UI
+ # can render preset pills with labels and blurbs without
+ # hardcoding any of it on the client side.
+ self._json({
+ "settings": get_settings(),
+ "presets": OUTPUT_PRESETS,
+ "default_preset": DEFAULT_OUTPUT_PRESET,
+ })
+ return
if parsed.path == "/file":
qs = parse_qs(parsed.query)
try:
@@ -2023,6 +2222,40 @@ class Handler(BaseHTTPRequestHandler):
HELPER.kill()
self._json({"ok": True}); return
+ if path == "/settings":
+ # Accept partial-patch updates: only the fields the user
+ # actually changed need to be present. Validation lives in
+ # _validate_settings_patch — never trust the form payload.
+ payload: dict = {}
+ for k, v in form.items():
+ payload[k] = v[0] if isinstance(v, list) else v
+ prev = get_settings()
+ current, err = update_settings(payload)
+ if err:
+ self._json({"ok": False, "error": err, "settings": current}, 400)
+ return
+ # Codec env vars are read at helper SPAWN time. If the user
+ # changed pix_fmt or crf, kill the helper so the next job
+ # respawns it with the new env. Job in flight finishes with
+ # the OLD codec settings (we're not interrupting a render).
+ codec_changed = (
+ prev.get("output_pix_fmt") != current.get("output_pix_fmt") or
+ prev.get("output_crf") != current.get("output_crf")
+ )
+ if codec_changed:
+ push(
+ f"settings: output codec → {current['output_pix_fmt']} "
+ f"crf {current['output_crf']} ({current['output_preset']}). "
+ f"Helper restarted; takes effect on next job."
+ )
+ HELPER.kill()
+ self._json({
+ "ok": True,
+ "settings": current,
+ "helper_restarted": codec_changed,
+ })
+ return
+
if path == "/stop":
stop_current_job()
self._json({"ok": True}); return
@@ -2700,6 +2933,80 @@ HTML = r"""<!doctype html>
.picker-recent-thumb:hover { border-color: var(--accent, #5a7cff); transform: translateY(-1px); }
.picker-recent-thumb.selected { border-color: var(--success, #3fb950); }
+ /* Header icon button — same height/feel as ghost-btn but icon-only.
+ Used for the settings cog. width:auto overrides the global
+ button{width:100%} rule. */
+ .icon-btn {
+ width: auto;
+ background: transparent; border: 1px solid var(--border);
+ color: var(--muted); padding: 4px 7px; border-radius: 6px;
+ cursor: pointer; display: inline-flex; align-items: center;
+ transition: border-color 120ms ease, color 120ms ease;
+ }
+ .icon-btn:hover { border-color: var(--accent); color: var(--accent-bright); }
+ .icon-btn svg { display: block; }
+
+ /* Settings modal — preset pills, advanced controls, Apply button.
+ Reuses the .models-modal frame so spacing/elevation match. */
+ .preset-grid {
+ display: grid; grid-template-columns: 1fr; gap: 8px;
+ margin: 12px 0;
+ }
+ .preset-card {
+ display: flex; align-items: flex-start; gap: 10px;
+ border: 1.5px solid var(--border); border-radius: 8px;
+ padding: 10px 12px; cursor: pointer; user-select: none;
+ transition: border-color 120ms ease, background 120ms ease;
+ background: var(--panel-2);
+ }
+ .preset-card:hover { border-color: var(--accent); }
+ .preset-card.active {
+ border-color: var(--accent);
+ background: var(--accent-dim, rgba(47,129,247,0.18));
+ }
+ .preset-card input[type="radio"] {
+ width: auto; margin: 4px 0 0 0; flex-shrink: 0;
+ accent-color: var(--accent);
+ }
+ .preset-card .preset-text { flex: 1; min-width: 0; }
+ .preset-card .preset-label {
+ font-size: 13px; font-weight: 600; color: var(--text);
+ margin: 0 0 4px 0;
+ }
+ .preset-card .preset-blurb {
+ font-size: 11px; color: var(--muted); line-height: 1.4;
+ }
+ .preset-card .preset-spec {
+ font-family: ui-monospace, SFMono-Regular, Menlo, monospace;
+ font-size: 10px; color: var(--accent-bright);
+ margin-top: 4px; letter-spacing: 0.02em;
+ }
+ .settings-section { margin-top: 14px; }
+ .settings-section h3 {
+ font-size: 11px; font-weight: 600; text-transform: uppercase;
+ letter-spacing: 0.08em; color: var(--muted); margin: 0 0 8px 0;
+ }
+ .settings-row { display: flex; gap: 10px; align-items: center; }
+ .settings-row label {
+ font-size: 11px; color: var(--muted); min-width: 70px;
+ }
+ .settings-foot {
+ display: flex; gap: 10px; justify-content: flex-end;
+ margin-top: 18px; padding-top: 14px;
+ border-top: 1px solid var(--border);
+ }
+ .settings-status { color: var(--muted); font-size: 11px; flex: 1; align-self: center; }
+ .settings-status.ok { color: var(--success, #3fb950); }
+ .settings-status.err { color: var(--danger, #f85149); }
+ button.primary-btn {
+ width: auto;
+ background: var(--accent); color: white; border: 1px solid var(--accent);
+ padding: 6px 14px; border-radius: 6px; font-size: 12px; font-weight: 500;
+ cursor: pointer;
+ }
+ button.primary-btn:hover { background: var(--accent-bright); }
+ button.primary-btn:disabled { opacity: 0.5; cursor: not-allowed; }
+
/* Models modal — opened by the header `models` pill. Layered on top
of everything; the form/log keep working underneath while a
download streams in (via the existing log panel at the bottom). */
@@ -2770,6 +3077,15 @@ HTML = r"""<!doctype html>
<span id="modelsPill" class="pill" style="cursor:pointer" onclick="openModelsModal()" title="View / download model status">models…</span>
<span id="queuePill" class="pill">queue 0</span>
<span id="jobPill" class="pill">idle</span>
+ <!-- Settings cog: opens the modal that lets users pick output codec
+ presets (Standard / Archival / Web / Custom). Sits between the
+ runtime pills and Stop Comfy so it's findable but not loud. -->
+ <button id="settingsBtn" class="icon-btn" onclick="openSettingsModal()" title="Output settings — codec, file size">
+ <svg viewBox="0 0 24 24" width="16" height="16" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round">
+ <circle cx="12" cy="12" r="3"></circle>
+ <path d="M19.4 15a1.65 1.65 0 0 0 .33 1.82l.06.06a2 2 0 0 1 0 2.83 2 2 0 0 1-2.83 0l-.06-.06a1.65 1.65 0 0 0-1.82-.33 1.65 1.65 0 0 0-1 1.51V21a2 2 0 0 1-4 0v-.09A1.65 1.65 0 0 0 9 19.4a1.65 1.65 0 0 0-1.82.33l-.06.06a2 2 0 0 1-2.83 0 2 2 0 0 1 0-2.83l.06-.06a1.65 1.65 0 0 0 .33-1.82 1.65 1.65 0 0 0-1.51-1H3a2 2 0 0 1 0-4h.09A1.65 1.65 0 0 0 4.6 9a1.65 1.65 0 0 0-.33-1.82l-.06-.06a2 2 0 0 1 0-2.83 2 2 0 0 1 2.83 0l.06.06a1.65 1.65 0 0 0 1.82.33H9a1.65 1.65 0 0 0 1-1.51V3a2 2 0 0 1 4 0v.09a1.65 1.65 0 0 0 1 1.51 1.65 1.65 0 0 0 1.82-.33l.06-.06a2 2 0 0 1 2.83 0 2 2 0 0 1 0 2.83l-.06.06a1.65 1.65 0 0 0-.33 1.82V9a1.65 1.65 0 0 0 1.51 1H21a2 2 0 0 1 0 4h-.09a1.65 1.65 0 0 0-1.51 1z"></path>
+ </svg>
+ </button>
<button id="stopComfyBtn" class="ghost-btn" style="display:none" onclick="api('/stop_comfy', 'POST').then(poll)">Stop Comfy</button>
<a class="creator-link" href="https://x.com/AIBizarrothe" target="_blank" rel="noopener" title="Follow Mr. Bizarro on X (the panel's creator)">
<img src="/assets/bizarro-avatar.jpg" class="creator-avatar" alt="">
@@ -3074,6 +3390,66 @@ HTML = r"""<!doctype html>
</section>
</main>
+<!-- ============== SETTINGS MODAL ============== -->
+<!-- Opened by the gear icon in the header. Lets users pick output codec
+ presets (Standard / Archival / Web) or set custom pix_fmt + crf.
+ Persisted to panel_settings.json; the helper subprocess restarts on
+ codec change so the new ffmpeg args take effect on next job. -->
+<div id="settingsModal" class="models-modal" style="display:none"
+ onclick="if(event.target===this) closeSettingsModal()">
+ <div class="models-card">
+ <div class="models-head">
+ <h2>Settings</h2>
+ <button class="ghost-btn" onclick="closeSettingsModal()">Close</button>
+ </div>
+ <div class="models-hint">
+ Pick how rendered clips are encoded. Settings are saved to
+ <code>panel_settings.json</code> and apply to every new render.
+ Files already in the gallery are not re-encoded.
+ </div>
+
+ <div class="settings-section">
+ <h3>Output format</h3>
+ <div class="preset-grid" id="settingsPresets">
+ <!-- populated by openSettingsModal() from /settings -->
+ </div>
+ </div>
+
+ <div class="settings-section" id="settingsCustomSection" style="display:none">
+ <h3>Custom (advanced)</h3>
+ <div class="settings-row" style="margin-bottom:10px">
+ <label>pix_fmt</label>
+ <select id="settingsPixFmt" style="flex:1">
+ <option value="yuv420p">yuv420p — web standard, broad support</option>
+ <option value="yuv422p">yuv422p — pro, more chroma than 420p</option>
+ <option value="yuv444p">yuv444p — full chroma, no subsampling</option>
+ <option value="yuv420p10le">yuv420p10le — 10-bit, HDR-ready</option>
+ <option value="yuv422p10le">yuv422p10le — 10-bit pro</option>
+ <option value="yuv444p10le">yuv444p10le — 10-bit, full chroma</option>
+ </select>
+ </div>
+ <div class="settings-row">
+ <label>CRF</label>
+ <input type="range" id="settingsCrfRange" min="0" max="30" step="1"
+ style="flex:1; width:auto" oninput="document.getElementById('settingsCrfNum').value = this.value">
+ <input type="number" id="settingsCrfNum" min="0" max="30" step="1"
+ style="width:60px"
+ oninput="document.getElementById('settingsCrfRange').value = this.value">
+ </div>
+ <div class="hint" style="margin-top:6px">
+ 0 = mathematically lossless · 18 = visually lossless ·
+ 23 = web default · 28+ = lossy
+ </div>
+ </div>
+
+ <div class="settings-foot">
+ <span class="settings-status" id="settingsStatus"></span>
+ <button class="ghost-btn" onclick="closeSettingsModal()">Cancel</button>
+ <button class="primary-btn" id="settingsApplyBtn" onclick="applySettings()">Apply</button>
+ </div>
+ </div>
+</div>
+
<!-- ============== TIER MODAL ============== -->
<!-- Opened by the "tier" pill in the header. Renders the detected
hardware tier + what it allows. Helps users understand WHY some
@@ -4287,6 +4663,125 @@ function openTierModal() {
}
function closeTierModal() { document.getElementById('tierModal').style.display = 'none'; }
+// ====== Settings modal ======
+// Single-shot fetch on open (settings change rarely). The modal hydrates
+// preset cards from the /settings response so the labels and blurbs
+// match the server-side OUTPUT_PRESETS table — no preset content
+// duplicated in JS.
+let _settingsCache = null;
+
+async function openSettingsModal() {
+ const modal = document.getElementById('settingsModal');
+ modal.style.display = 'flex';
+ document.getElementById('settingsStatus').textContent = '';
+ document.getElementById('settingsStatus').className = 'settings-status';
+ try {
+ const r = await fetch('/settings');
+ _settingsCache = await r.json();
+ } catch (e) {
+ document.getElementById('settingsStatus').textContent = 'Could not load settings.';
+ document.getElementById('settingsStatus').className = 'settings-status err';
+ return;
+ }
+ const cur = _settingsCache.settings;
+ const presets = _settingsCache.presets;
+ // Render preset cards (Standard, Archival, Web, Custom).
+ const order = ['standard', 'archival', 'web'];
+ const grid = document.getElementById('settingsPresets');
+ grid.innerHTML = '';
+ for (const key of order) {
+ const p = presets[key];
+ const active = cur.output_preset === key ? 'active' : '';
+ const card = document.createElement('label');
+ card.className = `preset-card ${active}`;
+ card.dataset.preset = key;
+ card.innerHTML = `
+ <input type="radio" name="settingsPreset" value="${key}" ${cur.output_preset === key ? 'checked' : ''}>
+ <div class="preset-text">
+ <div class="preset-label">${escapeHtml(p.label)}</div>
+ <div class="preset-blurb">${escapeHtml(p.blurb)}</div>
+ <div class="preset-spec">pix_fmt=${p.pix_fmt} · crf=${p.crf}</div>
+ </div>`;
+ card.addEventListener('click', () => selectPreset(key));
+ grid.appendChild(card);
+ }
+ // Custom row.
+ const customActive = cur.output_preset === 'custom' ? 'active' : '';
+ const custom = document.createElement('label');
+ custom.className = `preset-card ${customActive}`;
+ custom.dataset.preset = 'custom';
+ custom.innerHTML = `
+ <input type="radio" name="settingsPreset" value="custom" ${cur.output_preset === 'custom' ? 'checked' : ''}>
+ <div class="preset-text">
+ <div class="preset-label">Custom</div>
+ <div class="preset-blurb">Set pix_fmt and CRF manually. For unusual workflows (10-bit HDR, archival masters at non-standard CRF, format-specific delivery).</div>
+ <div class="preset-spec">pix_fmt=${cur.output_pix_fmt} · crf=${cur.output_crf}</div>
+ </div>`;
+ custom.addEventListener('click', () => selectPreset('custom'));
+ grid.appendChild(custom);
+ // Pre-fill custom inputs with current values
+ document.getElementById('settingsPixFmt').value = cur.output_pix_fmt;
+ document.getElementById('settingsCrfRange').value = cur.output_crf;
+ document.getElementById('settingsCrfNum').value = cur.output_crf;
+ document.getElementById('settingsCustomSection').style.display =
+ cur.output_preset === 'custom' ? 'block' : 'none';
+}
+
+function closeSettingsModal() {
+ document.getElementById('settingsModal').style.display = 'none';
+}
+
+function selectPreset(key) {
+ document.querySelectorAll('#settingsPresets .preset-card').forEach(c => {
+ c.classList.toggle('active', c.dataset.preset === key);
+ const r = c.querySelector('input[type="radio"]');
+ if (r) r.checked = (c.dataset.preset === key);
+ });
+ document.getElementById('settingsCustomSection').style.display =
+ key === 'custom' ? 'block' : 'none';
+ // Clear status so it doesn't claim "saved" after a fresh selection.
+ document.getElementById('settingsStatus').textContent = '';
+ document.getElementById('settingsStatus').className = 'settings-status';
+}
+
+async function applySettings() {
+ const status = document.getElementById('settingsStatus');
+ const btn = document.getElementById('settingsApplyBtn');
+ status.textContent = 'Saving…';
+ status.className = 'settings-status';
+ btn.disabled = true;
+ // Read which preset is selected. Custom path also sends pix_fmt + crf.
+ const checked = document.querySelector('#settingsPresets input[type="radio"]:checked');
+ const preset = checked ? checked.value : 'standard';
+ const fd = new FormData();
+ fd.set('output_preset', preset);
+ if (preset === 'custom') {
+ fd.set('output_pix_fmt', document.getElementById('settingsPixFmt').value);
+ fd.set('output_crf', document.getElementById('settingsCrfNum').value);
+ }
+ try {
+ const r = await fetch('/settings', { method: 'POST', body: fd });
+ const data = await r.json();
+ if (!r.ok || data.error) {
+ status.textContent = data.error || `HTTP ${r.status}`;
+ status.className = 'settings-status err';
+ btn.disabled = false;
+ return;
+ }
+ status.textContent = data.helper_restarted
+ ? 'Saved. Helper restarted — takes effect on the next render.'
+ : 'Saved.';
+ status.className = 'settings-status ok';
+ btn.disabled = false;
+ // Refresh cache so a re-open shows the new values without a stale flash.
+ _settingsCache = { ...(_settingsCache || {}), settings: data.settings };
+ } catch (e) {
+ status.textContent = 'Network error: ' + (e.message || e);
+ status.className = 'settings-status err';
+ btn.disabled = false;
+ }
+}
+
// ====== Models modal ======
// Opens to /models snapshot. While open, the main poll() refreshes the
// list every cycle so download progress appears live. Each row shows: