Hi — we run a small studio producing branded vertical (1080×1920) real-estate reels: same template, different listing data + brand kit per client, aiming for hundreds/month. We drive Palmier Pro entirely through the local MCP with a deterministic script (no LLM in the loop), and a prototype that composes + exports a full branded reel headless already works great. Before committing deeper, some questions only you can answer:
1. Unattended / headless operation. Can the MCP server drive a full build→export unattended — i.e. with the Palmier window backgrounded or the Mac left running headless — or does the GUI need to be foregrounded/interacted-with for generation, render, or export to proceed reliably?
2. Parallelism / throughput. Is it one active project + one active timeline per app instance (serial), or is there any supported way to run builds in parallel — multiple instances, a queue, or a batch/render-farm mode? What render throughput do you consider realistic on one Apple Silicon Mac for many short reels/day?
3. Programmatic batch / template-merge on the roadmap? Is there (or planned) a first-class way to render N videos from a saved template by merging a data table — a REST API, CLI, or "brand template + variables" feature — versus scripting the localhost MCP call-by-call?
4. MCP surface stability. You ship multiple 0.x releases/week. What's your commitment to stability of the localhost MCP tool contract (tool names, params, semantics) across releases — i.e. can we build automation on it without it breaking on updates? Any versioning/pinning?
5. Bug report — manage_tracks (v0.6.4). In our prototype, manage_tracks reorder was a silent no-op across several calls (including a trivial index 1→0 swap — returned a success-shaped result but track order never changed), and a remove by index deleted a different track than the one at that index (removed our text track instead of the targeted top image track). Track labels also appeared to shift/reassign confusingly after edits. For unattended automation, silent mis-targeting on track ops is a blocker — is this a known issue, and is correct index-addressing for reorder/remove something you can confirm/fix?
6. Auth surface. The MCP endpoint at 127.0.0.1:19789/mcp answers with no authentication. Is that intended for localhost-only, and are there plans for a token so a script can't be driven by any local process?
7. Text animation depth. Branded titles are core to our reels. Today add_texts offers a fixed preset list (fadeIn, wordReveal, popIn, typewriter…). Three things we'd need: (a) custom fonts — does fontName accept any installed OTF/TTF, and can a project bundle a font so renders are portable? (b) designable motion — any path to custom easing, per-word timing, or keyframeable text animation beyond the presets? (c) motion-matched text — is there (or planned) point/camera tracking so a title can be pinned to the scene and move with a dolly shot, rather than sitting static on top?
8. Generated music in client advertising. We'd use ElevenLabs Music / Sonilo tracks as beds in reels our clients run as paid ads (Meta placements). Your terms allow commercial use of generated content — does that explicitly cover third-party advertising use by our clients, and does the underlying model provider's license (ElevenLabs/Sonilo) flow through Palmier for that? Any indemnification if a generated track is claimed as substantially similar to an existing work (Content ID claims on ads are a real operational risk)?
Happy to share the prototype driver script or repro details for #5. Congrats on the launch — the local-MCP architecture is genuinely great to build against.
Hi — we run a small studio producing branded vertical (1080×1920) real-estate reels: same template, different listing data + brand kit per client, aiming for hundreds/month. We drive Palmier Pro entirely through the local MCP with a deterministic script (no LLM in the loop), and a prototype that composes + exports a full branded reel headless already works great. Before committing deeper, some questions only you can answer:
1. Unattended / headless operation. Can the MCP server drive a full build→export unattended — i.e. with the Palmier window backgrounded or the Mac left running headless — or does the GUI need to be foregrounded/interacted-with for generation, render, or export to proceed reliably?
2. Parallelism / throughput. Is it one active project + one active timeline per app instance (serial), or is there any supported way to run builds in parallel — multiple instances, a queue, or a batch/render-farm mode? What render throughput do you consider realistic on one Apple Silicon Mac for many short reels/day?
3. Programmatic batch / template-merge on the roadmap? Is there (or planned) a first-class way to render N videos from a saved template by merging a data table — a REST API, CLI, or "brand template + variables" feature — versus scripting the localhost MCP call-by-call?
4. MCP surface stability. You ship multiple 0.x releases/week. What's your commitment to stability of the localhost MCP tool contract (tool names, params, semantics) across releases — i.e. can we build automation on it without it breaking on updates? Any versioning/pinning?
5. Bug report —
manage_tracks(v0.6.4). In our prototype,manage_tracksreorderwas a silent no-op across several calls (including a trivial index 1→0 swap — returned a success-shaped result but track order never changed), and aremoveby index deleted a different track than the one at that index (removed our text track instead of the targeted top image track). Track labels also appeared to shift/reassign confusingly after edits. For unattended automation, silent mis-targeting on track ops is a blocker — is this a known issue, and is correct index-addressing for reorder/remove something you can confirm/fix?6. Auth surface. The MCP endpoint at
127.0.0.1:19789/mcpanswers with no authentication. Is that intended for localhost-only, and are there plans for a token so a script can't be driven by any local process?7. Text animation depth. Branded titles are core to our reels. Today
add_textsoffers a fixed preset list (fadeIn, wordReveal, popIn, typewriter…). Three things we'd need: (a) custom fonts — doesfontNameaccept any installed OTF/TTF, and can a project bundle a font so renders are portable? (b) designable motion — any path to custom easing, per-word timing, or keyframeable text animation beyond the presets? (c) motion-matched text — is there (or planned) point/camera tracking so a title can be pinned to the scene and move with a dolly shot, rather than sitting static on top?8. Generated music in client advertising. We'd use ElevenLabs Music / Sonilo tracks as beds in reels our clients run as paid ads (Meta placements). Your terms allow commercial use of generated content — does that explicitly cover third-party advertising use by our clients, and does the underlying model provider's license (ElevenLabs/Sonilo) flow through Palmier for that? Any indemnification if a generated track is claimed as substantially similar to an existing work (Content ID claims on ads are a real operational risk)?
Happy to share the prototype driver script or repro details for #5. Congrats on the launch — the local-MCP architecture is genuinely great to build against.