@gobi-ai/cli 0.9.9 → 0.9.10

package/.claude-plugin/marketplace.json

@@ -4,12 +4,12 @@
     "name": "gobi-ai"
   },
   "description": "Claude Code plugin for the Gobi collaborative knowledge platform CLI",
-  "version": "0.9.9",
+  "version": "0.9.10",
   "plugins": [
     {
       "name": "gobi",
       "description": "Manage the Gobi collaborative knowledge platform from the command line. Search and ask brains, publish brain documents, create threads, manage sessions, generate images and videos.",
-      "version": "0.9.9",
+      "version": "0.9.10",
       "author": {
         "name": "gobi-ai"
       },

package/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "gobi",
   "description": "Manage the Gobi collaborative knowledge platform from the command line",
-  "version": "0.9.9",
+  "version": "0.9.10",
   "author": {
     "name": "gobi-ai"
   },

package/dist/commands/media.js

@@ -61,6 +61,47 @@ async function downloadVideoToFile(videoId, outputPath) {
     await writeFile(outputPath, buffer);
     return { contentType: ct || "video/mp4", size: buffer.length };
 }
+const MIME_MAP = {
+    ".png": "image/png",
+    ".jpg": "image/jpeg",
+    ".jpeg": "image/jpeg",
+    ".webp": "image/webp",
+    ".gif": "image/gif",
+    ".mp4": "video/mp4",
+    ".mov": "video/quicktime",
+    ".mp3": "audio/mpeg",
+    ".wav": "audio/wav",
+};
+/**
+ * Upload a local file and return its media ID.
+ * Handles init → PUT → finalize in one call.
+ */
+async function uploadFile(filePath) {
+    const { readFile, stat } = await import("fs/promises");
+    const { basename, extname } = await import("path");
+    const buffer = await readFile(filePath);
+    const fileName = basename(filePath);
+    const fileSize = (await stat(filePath)).size;
+    const ext = extname(filePath).toLowerCase();
+    const contentType = MIME_MAP[ext] || "application/octet-stream";
+    const initResp = (await apiPost("/media-gen/media/initialize", {
+        fileName, contentType, fileSize,
+    }));
+    const initData = unwrapResp(initResp);
+    const mediaId = initData.mediaId;
+    const uploadUrl = initData.uploadUrl;
+    if (!mediaId || !uploadUrl)
+        throw new Error("Upload init failed: missing mediaId or uploadUrl");
+    const putRes = await fetch(uploadUrl, {
+        method: "PUT",
+        headers: { "Content-Type": contentType },
+        body: buffer,
+    });
+    if (!putRes.ok)
+        throw new Error(`Upload PUT failed: ${putRes.status}`);
+    await apiPost("/media-gen/media/finalize", { mediaId });
+    return mediaId;
+}
 function extractImageUrl(data) {
     return (data.downloadUrl || data.download_url || data.url);
 }
@@ -72,44 +113,15 @@ export function registerMediaCommand(program) {
     //  Upload
     // ════════════════════════════════════════════════════════════════════
     media
-        .command("upload-init")
-        .description("Get a presigned upload URL for a media file.")
-        .requiredOption("--file-name <fileName>", "Name of the file to upload")
-        .requiredOption("--content-type <contentType>", "MIME type (e.g. image/png, video/mp4)")
-        .option("--file-size <fileSize>", "File size in bytes")
-        .action(async (opts) => {
-        const body = {
-            fileName: opts.fileName,
-            contentType: opts.contentType,
-        };
-        if (opts.fileSize)
-            body.fileSize = parseInt(opts.fileSize, 10);
-        const resp = (await apiPost("/media-gen/media/initialize", body));
-        const data = unwrapResp(resp);
-        if (isJsonMode(media)) {
-            jsonOut(data);
-            return;
-        }
-        console.log(`Upload initialized!\n` +
-            `  Media ID:   ${data.mediaId}\n` +
-            `  Upload URL: ${data.uploadUrl}\n\n` +
-            `PUT your file to the upload URL, then run:\n` +
-            `  gobi media upload-finalize --media-id ${data.mediaId}`);
-    });
-    media
-        .command("upload-finalize")
-        .description("Confirm that a media upload is complete.")
-        .requiredOption("--media-id <mediaId>", "Media ID from upload-init")
-        .action(async (opts) => {
-        const resp = (await apiPost("/media-gen/media/finalize", {
-            mediaId: opts.mediaId,
-        }));
-        const data = unwrapResp(resp);
+        .command("upload <file>")
+        .description("Upload a local file and return its media ID.")
+        .action(async (file) => {
+        const mediaId = await uploadFile(file);
         if (isJsonMode(media)) {
-            jsonOut(data);
+            jsonOut({ mediaId });
             return;
         }
-        console.log(`Upload finalized for media ${opts.mediaId}.`);
+        console.log(`Uploaded → Media ID: ${mediaId}`);
     });
     // ════════════════════════════════════════════════════════════════════
     //  Avatars & Voices
@@ -162,7 +174,7 @@ export function registerMediaCommand(program) {
         .requiredOption("--avatar-id <avatarId>", "Avatar to use")
         .requiredOption("--voice-id <voiceId>", "Voice to use")
         .requiredOption("--script <script>", "Script for the avatar to read")
-        .option("--background-media-id <backgroundMediaId>", "Background media ID (from upload)")
+        .option("--background <file>", "Background image file (auto-uploaded)")
         .option("--wait", "Poll until generation completes")
         .option("-o, --output <path>", "Download video to this path when done (implies --wait)")
         .action(async (opts) => {
@@ -174,8 +186,8 @@ export function registerMediaCommand(program) {
             voiceId: opts.voiceId,
             script: opts.script,
         };
-        if (opts.backgroundMediaId)
-            body.backgroundMediaId = opts.backgroundMediaId;
+        if (opts.background)
+            body.backgroundMediaId = await uploadFile(opts.background);
         const resp = (await apiPost("/media-gen/videos", body));
         let data = unwrapResp(resp);
         const videoId = data.id || data.videoId || data.jobId;
@@ -336,9 +348,9 @@ export function registerMediaCommand(program) {
         .option("--generate-audio", "Generate audio for the video")
         .option("--negative-prompt <negativePrompt>", "Negative prompt")
         .option("--sample-count <count>", "Number of samples (1-4)")
-        .option("--first-frame-media-id <mediaId>", "First frame image media ID")
-        .option("--last-frame-media-id <mediaId>", "Last frame image media ID")
-        .option("--reference-media-ids <ids>", "Comma-separated reference image media IDs (max 3)")
+        .option("--first-frame <file>", "First frame image file (auto-uploaded)")
+        .option("--last-frame <file>", "Last frame image file (auto-uploaded)")
+        .option("--reference-images <files>", "Comma-separated reference image files (auto-uploaded, max 3)")
         .option("--wait", "Poll until generation completes")
         .option("-o, --output <path>", "Download video to this path when done (implies --wait)")
         .action(async (opts) => {
@@ -370,12 +382,14 @@ export function registerMediaCommand(program) {
                 throw new Error("--sample-count must be a number");
             body.sampleCount = v;
         }
-        if (opts.firstFrameMediaId)
-            body.firstFrameImageMediaId = opts.firstFrameMediaId;
-        if (opts.lastFrameMediaId)
-            body.lastFrameImageMediaId = opts.lastFrameMediaId;
-        if (opts.referenceMediaIds)
-            body.referenceImageMediaIds = opts.referenceMediaIds.split(",").map((s) => s.trim());
+        if (opts.firstFrame)
+            body.firstFrameImageMediaId = await uploadFile(opts.firstFrame);
+        if (opts.lastFrame)
+            body.lastFrameImageMediaId = await uploadFile(opts.lastFrame);
+        if (opts.referenceImages) {
+            const files = opts.referenceImages.split(",").map((s) => s.trim());
+            body.referenceImageMediaIds = await Promise.all(files.map((f) => uploadFile(f)));
+        }
         const resp = (await apiPost("/media-gen/videos/cinematic", body));
         let data = unwrapResp(resp);
         const videoId = data.id || data.videoId || data.jobId;
@@ -413,18 +427,18 @@ export function registerMediaCommand(program) {
     media
         .command("avatar-design")
         .description("Start a design-your-avatar job.")
-        .requiredOption("--name <name>", "Name for the avatar")
+        .option("--name <name>", "Name for the avatar (auto-generated if omitted)")
         .requiredOption("--gender <gender>", "Gender for the avatar design")
         .requiredOption("--age <age>", "Age range for the avatar")
         .requiredOption("--ethnicity <ethnicity>", "Ethnicity for the avatar")
         .requiredOption("--outfit <outfit>", "Outfit description")
         .requiredOption("--background <background>", "Background description")
         .option("--no-portrait", "Generate full-body instead of portrait")
-        .option("--audio-media-id <mediaId>", "Custom voice audio media ID")
+        .option("--audio <file>", "Custom voice audio file (auto-uploaded)")
         .option("--wait", "Poll until variants are ready")
         .action(async (opts) => {
         const body = {
-            name: opts.name,
+            name: opts.name || `avatar-${new Date().toISOString().replace(/[:.]/g, "-").slice(0, 19)}`,
             gender: opts.gender,
             age: opts.age,
             ethnicity: opts.ethnicity,
@@ -432,8 +446,8 @@ export function registerMediaCommand(program) {
             background: opts.background,
             isPortrait: opts.portrait,
         };
-        if (opts.audioMediaId)
-            body.audioMediaId = opts.audioMediaId;
+        if (opts.audio)
+            body.audioMediaId = await uploadFile(opts.audio);
         const resp = (await apiPost("/media-gen/avatars/design", body));
         let data = unwrapResp(resp);
         const jobId = data.jobId || data.id;
@@ -479,20 +493,20 @@ export function registerMediaCommand(program) {
     media
         .command("avatar-from-selfie")
         .description("Create an avatar from a selfie (instant or enhanced with prompt).")
-        .requiredOption("--name <name>", "Name for the avatar")
-        .requiredOption("--photo-media-id <mediaId>", "Selfie photo media ID")
+        .option("--name <name>", "Name for the avatar (auto-generated if omitted)")
+        .requiredOption("--photo <file>", "Selfie photo file (auto-uploaded)")
         .option("--prompt <prompt>", "Enhancement prompt (triggers async enhance flow)")
-        .option("--audio-media-id <mediaId>", "Custom voice audio media ID")
+        .option("--audio <file>", "Custom voice audio file (auto-uploaded)")
         .option("--wait", "Poll until job completes (only for enhance flow)")
         .action(async (opts) => {
         const body = {
-            name: opts.name,
-            photoMediaId: opts.photoMediaId,
+            name: opts.name || `avatar-${new Date().toISOString().replace(/[:.]/g, "-").slice(0, 19)}`,
+            photoMediaId: await uploadFile(opts.photo),
         };
         if (opts.prompt)
             body.prompt = opts.prompt;
-        if (opts.audioMediaId)
-            body.audioMediaId = opts.audioMediaId;
+        if (opts.audio)
+            body.audioMediaId = await uploadFile(opts.audio);
         const resp = (await apiPost("/media-gen/avatars/from-selfie", body));
         let data = unwrapResp(resp);
         const jobId = data.jobId || data.id;
@@ -548,7 +562,7 @@ export function registerMediaCommand(program) {
         .option("--aspect-ratio <aspectRatio>", "Aspect ratio (1:1, 16:9, 9:16, 4:3, 3:4)")
         .option("--negative-prompt <negativePrompt>", "Negative prompt")
         .option("--seed <seed>", "Random seed for reproducibility")
-        .option("--reference-media-id <referenceMediaId>", "Reference image media ID")
+        .option("--reference-image <file>", "Reference image file (auto-uploaded)")
         .option("--wait", "Poll until generation completes")
         .option("-o, --output <path>", "Download image to this path when done (implies --wait)")
         .action(async (opts) => {
@@ -566,8 +580,8 @@ export function registerMediaCommand(program) {
             body.negativePrompt = opts.negativePrompt;
         if (opts.seed)
             body.seed = parseInt(opts.seed, 10);
-        if (opts.referenceMediaId)
-            body.referenceMediaId = opts.referenceMediaId;
+        if (opts.referenceImage)
+            body.referenceMediaId = await uploadFile(opts.referenceImage);
         const resp = (await apiPost("/media-gen/images/generate", body));
         let data = unwrapResp(resp);
         const jobId = data.jobId || data.id;
@@ -624,19 +638,23 @@ export function registerMediaCommand(program) {
     media
         .command("image-edit")
         .description("Edit an existing image with a prompt (image-to-image).")
-        .requiredOption("--media-id <mediaId>", "Source image media ID")
+        .requiredOption("--image <file>", "Source image file (auto-uploaded)")
         .requiredOption("--prompt <prompt>", "Edit instruction")
-        .requiredOption("--name <name>", "Name for the edited image")
+        .option("--name <name>", "Name for the edited image (auto-generated if omitted)")
         .option("--wait", "Poll until generation completes")
+        .option("-o, --output <path>", "Download image to this path when done (implies --wait)")
         .action(async (opts) => {
+        const shouldWait = opts.wait || !!opts.output;
+        const mediaId = await uploadFile(opts.image);
+        const autoName = opts.name || opts.prompt.slice(0, 50).replace(/[^a-zA-Z0-9-_ ]/g, "").trim().replace(/\s+/g, "-");
         const resp = (await apiPost("/media-gen/images/edit", {
-            mediaId: opts.mediaId,
+            mediaId,
             prompt: opts.prompt,
-            name: opts.name,
+            name: autoName,
         }));
         let data = unwrapResp(resp);
         const jobId = data.jobId || data.id;
-        if (opts.wait && jobId) {
+        if (shouldWait && jobId) {
             console.log(`Image edit job ${jobId} — polling for completion…`);
             data = await pollStatus(`/media-gen/images/${jobId}`, [
                 "completed",
@@ -645,6 +663,29 @@ export function registerMediaCommand(program) {
                 "inference_failed",
             ]);
         }
+        // Download image to file if -o specified
+        if (opts.output && jobId) {
+            const token = await getValidToken();
+            const query = "";
+            const url = `${BASE_URL}/media-gen/images/${jobId}/download${query}`;
+            const res = await fetch(url, {
+                headers: { Authorization: `Bearer ${token}` },
+            });
+            if (res.ok) {
+                const { writeFile, mkdir } = await import("fs/promises");
+                const { dirname } = await import("path");
+                const buffer = Buffer.from(await res.arrayBuffer());
+                await mkdir(dirname(opts.output), { recursive: true });
+                await writeFile(opts.output, buffer);
+                const contentType = res.headers.get("content-type") || "image/png";
+                if (isJsonMode(media)) {
+                    jsonOut({ ...data, filename: opts.output, contentType, size: buffer.length });
+                    return;
+                }
+                console.log(`Image saved to ${opts.output} (${buffer.length} bytes)`);
+                return;
+            }
+        }
         if (isJsonMode(media)) {
             jsonOut(data);
             return;
@@ -660,21 +701,28 @@ export function registerMediaCommand(program) {
     media
         .command("image-inpaint")
         .description("Inpaint an image region using a mask.")
-        .requiredOption("--media-id <mediaId>", "Source image media ID")
-        .requiredOption("--mask-media-id <maskMediaId>", "Mask image media ID")
+        .requiredOption("--image <file>", "Source image file (auto-uploaded)")
+        .requiredOption("--mask <file>", "Mask image file (auto-uploaded)")
         .requiredOption("--prompt <prompt>", "Inpainting prompt")
-        .requiredOption("--name <name>", "Name for the inpainted image")
+        .option("--name <name>", "Name for the inpainted image (auto-generated if omitted)")
         .option("--wait", "Poll until generation completes")
+        .option("-o, --output <path>", "Download image to this path when done (implies --wait)")
         .action(async (opts) => {
+        const shouldWait = opts.wait || !!opts.output;
+        const [mediaId, maskMediaId] = await Promise.all([
+            uploadFile(opts.image),
+            uploadFile(opts.mask),
+        ]);
+        const autoName = opts.name || opts.prompt.slice(0, 50).replace(/[^a-zA-Z0-9-_ ]/g, "").trim().replace(/\s+/g, "-");
         const resp = (await apiPost("/media-gen/images/inpaint", {
-            mediaId: opts.mediaId,
-            maskMediaId: opts.maskMediaId,
+            mediaId,
+            maskMediaId,
             prompt: opts.prompt,
-            name: opts.name,
+            name: autoName,
         }));
         let data = unwrapResp(resp);
         const jobId = data.jobId || data.id;
-        if (opts.wait && jobId) {
+        if (shouldWait && jobId) {
             console.log(`Inpaint job ${jobId} — polling for completion…`);
             data = await pollStatus(`/media-gen/images/${jobId}`, [
                 "completed",
@@ -683,6 +731,28 @@ export function registerMediaCommand(program) {
                 "inference_failed",
             ]);
         }
+        // Download image to file if -o specified
+        if (opts.output && jobId) {
+            const token = await getValidToken();
+            const url = `${BASE_URL}/media-gen/images/${jobId}/download`;
+            const res = await fetch(url, {
+                headers: { Authorization: `Bearer ${token}` },
+            });
+            if (res.ok) {
+                const { writeFile, mkdir } = await import("fs/promises");
+                const { dirname } = await import("path");
+                const buffer = Buffer.from(await res.arrayBuffer());
+                await mkdir(dirname(opts.output), { recursive: true });
+                await writeFile(opts.output, buffer);
+                const contentType = res.headers.get("content-type") || "image/png";
+                if (isJsonMode(media)) {
+                    jsonOut({ ...data, filename: opts.output, contentType, size: buffer.length });
+                    return;
+                }
+                console.log(`Image saved to ${opts.output} (${buffer.length} bytes)`);
+                return;
+            }
+        }
         if (isJsonMode(media)) {
             jsonOut(data);
             return;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gobi-ai/cli",
-  "version": "0.9.9",
+  "version": "0.9.10",
   "description": "CLI client for the Gobi collaborative knowledge platform",
   "license": "MIT",
   "type": "module",

package/skills/gobi-media/SKILL.md

@@ -67,10 +67,10 @@ gobi --json media video-create --avatar-id "<AVATAR_ID>" --voice-id "<VOICE_ID>"
 
 The `-o` flag implies `--wait` and downloads the video when done.
 
-To use a custom image as the **background** of a video, upload it via `upload-init` / `upload-finalize`, then pass the mediaId as `--background-media-id`:
+To use a custom image as the **background** of a video, pass it directly as `--background <file>` (auto-uploaded):
 
 ```bash
-gobi --json media video-create --avatar-id "<AVATAR_ID>" --voice-id "<VOICE_ID>" --script "<SCRIPT>" --background-media-id "<MEDIA_ID>" -o media/<NAME>.mp4
+gobi --json media video-create --avatar-id "<AVATAR_ID>" --voice-id "<VOICE_ID>" --script "<SCRIPT>" --background media/bg.png -o media/<NAME>.mp4
 ```
 
 ## Typical Workflow (Cinematic Video)
@@ -81,7 +81,17 @@ Generate a cinematic video from a text prompt (no avatar needed):
 gobi --json media cinematic-create --prompt "<PROMPT>" --aspect-ratio "<RATIO>" -o media/<NAME>.mp4
 ```
 
-Options: `--duration <4-8>`, `--resolution <720p|1080p>`, `--enhance-prompt`, `--generate-audio`, `--negative-prompt`, `--sample-count <1-4>`, `--first-frame-media-id`, `--last-frame-media-id`, `--reference-media-ids`.
+Options: `--duration <4-8>`, `--resolution <720p|1080p>`, `--enhance-prompt`, `--generate-audio`, `--negative-prompt`, `--sample-count <1-4>`, `--first-frame <file>`, `--last-frame <file>`, `--reference-images <files>`.
+
+## Typical Workflow (Image Editing)
+
+Edit an existing image with a prompt — single command:
+
+```bash
+gobi --json media image-edit --image media/source.png --prompt "<EDIT_INSTRUCTION>" -o media/<NAME>.png
+```
+
+All file arguments (`--image`, `--mask`, `--background`, `--photo`, `--audio`, `--reference-image`, `--first-frame`, `--last-frame`) accept local file paths and auto-upload them. No need to manually upload first.
 
 ## Custom Avatars
 
@@ -90,7 +100,7 @@ Three ways to create custom avatars:
 ### 1. Design from scratch
 
 ```bash
-gobi --json media avatar-design --name "<NAME>" --gender "<GENDER>" --age "<AGE>" --ethnicity "<ETHNICITY>" --outfit "<OUTFIT>" --background "<BACKGROUND>" --wait
+gobi --json media avatar-design --gender "<GENDER>" --age "<AGE>" --ethnicity "<ETHNICITY>" --outfit "<OUTFIT>" --background "<BACKGROUND>" --wait
 ```
 
 When `variants_ready`, confirm with:
@@ -102,15 +112,13 @@ gobi --json media avatar-confirm --job-id "<JOB_ID>"
 ### 2. From a selfie (instant)
 
 ```bash
-gobi --json media avatar-from-selfie --name "<NAME>" --photo-media-id "<MEDIA_ID>"
+gobi --json media avatar-from-selfie --photo media/selfie.png
 ```
 
-Upload the selfie first via `upload-init` / `upload-finalize` to get the media ID.
-
 ### 3. From a selfie (enhanced with prompt)
 
 ```bash
-gobi --json media avatar-from-selfie --name "<NAME>" --photo-media-id "<MEDIA_ID>" --prompt "<ENHANCEMENT>" --wait
+gobi --json media avatar-from-selfie --photo media/selfie.png --prompt "<ENHANCEMENT>" --wait
 ```
 
 Check any avatar job status with:
@@ -131,8 +139,7 @@ Do NOT use markdown image/link syntax `![](...)` or `gobi://` URLs. Always use `
 
 ### Upload
 
-- `gobi media upload-init` — Get a presigned upload URL for a media file.
-- `gobi media upload-finalize` — Confirm that a media upload is complete.
+- `gobi media upload <file>` — Upload a local file and return its media ID. Content type is auto-detected.
 
 ### Avatars & Voices
 
@@ -161,6 +168,8 @@ Do NOT use markdown image/link syntax `![](...)` or `gobi://` URLs. Always use `
 - `gobi media image-edit` — Edit an existing image with a prompt (image-to-image).
 - `gobi media image-inpaint` — Inpaint an image region using a mask.
 - `gobi media image-status` — Check image generation job status.
+- `gobi media image-download` — Download a generated image.
+- `gobi media image-status` — Check image generation job status.
 
 ## Reference Documentation
 

package/skills/gobi-media/references/media.md

@@ -9,8 +9,7 @@ Options:
   -h, --help                           display help for command
 
 Commands:
-  upload-init [options]                Get a presigned upload URL for a media file.
-  upload-finalize [options]            Confirm that a media upload is complete.
+  upload <file>                        Upload a local file and return its media ID.
   avatars                              List available avatars.
   voices                               List available voices.
   video-create [options]               Create an avatar video generation job.
@@ -31,30 +30,15 @@ Commands:
   help [command]                       display help for command
 ```
 
-## upload-init
+## upload
 
 ```
-Usage: gobi media upload-init [options]
+Usage: gobi media upload [options] <file>
 
-Get a presigned upload URL for a media file.
+Upload a local file and return its media ID.
 
 Options:
-  --file-name <fileName>        Name of the file to upload
-  --content-type <contentType>  MIME type (e.g. image/png, video/mp4)
-  --file-size <fileSize>        File size in bytes
-  -h, --help                    display help for command
-```
-
-## upload-finalize
-
-```
-Usage: gobi media upload-finalize [options]
-
-Confirm that a media upload is complete.
-
-Options:
-  --media-id <mediaId>  Media ID from upload-init
-  -h, --help            display help for command
+  -h, --help  display help for command
 ```
 
 ## avatars
@@ -87,14 +71,14 @@ Usage: gobi media video-create [options]
 Create an avatar video generation job.
 
 Options:
-  --name <name>                              Name for the video (auto-generated if omitted)
-  --avatar-id <avatarId>                     Avatar to use
-  --voice-id <voiceId>                       Voice to use
-  --script <script>                          Script for the avatar to read
-  --background-media-id <backgroundMediaId>  Background media ID (from upload)
-  --wait                                     Poll until generation completes
-  -o, --output <path>                        Download video to this path when done (implies --wait)
-  -h, --help                                 display help for command
+  --name <name>           Name for the video (auto-generated if omitted)
+  --avatar-id <avatarId>  Avatar to use
+  --voice-id <voiceId>    Voice to use
+  --script <script>       Script for the avatar to read
+  --background <file>     Background image file (auto-uploaded)
+  --wait                  Poll until generation completes
+  -o, --output <path>     Download video to this path when done (implies --wait)
+  -h, --help              display help for command
 ```
 
 ## video-list
@@ -161,9 +145,9 @@ Options:
   --generate-audio                    Generate audio for the video
   --negative-prompt <negativePrompt>  Negative prompt
   --sample-count <count>              Number of samples (1-4)
-  --first-frame-media-id <mediaId>    First frame image media ID
-  --last-frame-media-id <mediaId>     Last frame image media ID
-  --reference-media-ids <ids>         Comma-separated reference image media IDs (max 3)
+  --first-frame <file>                First frame image file (auto-uploaded)
+  --last-frame <file>                 Last frame image file (auto-uploaded)
+  --reference-images <files>          Comma-separated reference image files (auto-uploaded, max 3)
   --wait                              Poll until generation completes
   -o, --output <path>                 Download video to this path when done (implies --wait)
   -h, --help                          display help for command
@@ -177,16 +161,16 @@ Usage: gobi media avatar-design [options]
 Start a design-your-avatar job.
 
 Options:
-  --name <name>               Name for the avatar
-  --gender <gender>           Gender for the avatar design
-  --age <age>                 Age range for the avatar
-  --ethnicity <ethnicity>     Ethnicity for the avatar
-  --outfit <outfit>           Outfit description
-  --background <background>   Background description
-  --no-portrait               Generate full-body instead of portrait
-  --audio-media-id <mediaId>  Custom voice audio media ID
-  --wait                      Poll until variants are ready
-  -h, --help                  display help for command
+  --name <name>              Name for the avatar (auto-generated if omitted)
+  --gender <gender>          Gender for the avatar design
+  --age <age>                Age range for the avatar
+  --ethnicity <ethnicity>    Ethnicity for the avatar
+  --outfit <outfit>          Outfit description
+  --background <background>  Background description
+  --no-portrait              Generate full-body instead of portrait
+  --audio <file>             Custom voice audio file (auto-uploaded)
+  --wait                     Poll until variants are ready
+  -h, --help                 display help for command
 ```
 
 ## avatar-confirm
@@ -210,12 +194,12 @@ Usage: gobi media avatar-from-selfie [options]
 Create an avatar from a selfie (instant or enhanced with prompt).
 
 Options:
-  --name <name>               Name for the avatar
-  --photo-media-id <mediaId>  Selfie photo media ID
-  --prompt <prompt>           Enhancement prompt (triggers async enhance flow)
-  --audio-media-id <mediaId>  Custom voice audio media ID
-  --wait                      Poll until job completes (only for enhance flow)
-  -h, --help                  display help for command
+  --name <name>      Name for the avatar (auto-generated if omitted)
+  --photo <file>     Selfie photo file (auto-uploaded)
+  --prompt <prompt>  Enhancement prompt (triggers async enhance flow)
+  --audio <file>     Custom voice audio file (auto-uploaded)
+  --wait             Poll until job completes (only for enhance flow)
+  -h, --help         display help for command
 ```
 
 ## avatar-job-status
@@ -238,16 +222,16 @@ Usage: gobi media image-generate [options]
 Generate an image from a text prompt. Types: image (default), thumbnail (YouTube-optimized), asset (logo/product). Aspect ratios: 1:1, 16:9, 9:16, 4:3, 3:4
 
 Options:
-  --prompt <prompt>                        Text prompt for image generation
-  --name <name>                            Name for the generated image (auto-generated from prompt if omitted)
-  --type <type>                            Generation type: image (default), thumbnail (YouTube-optimized), asset (logo/product)
-  --aspect-ratio <aspectRatio>             Aspect ratio (1:1, 16:9, 9:16, 4:3, 3:4)
-  --negative-prompt <negativePrompt>       Negative prompt
-  --seed <seed>                            Random seed for reproducibility
-  --reference-media-id <referenceMediaId>  Reference image media ID
-  --wait                                   Poll until generation completes
-  -o, --output <path>                      Download image to this path when done (implies --wait)
-  -h, --help                               display help for command
+  --prompt <prompt>                   Text prompt for image generation
+  --name <name>                       Name for the generated image (auto-generated from prompt if omitted)
+  --type <type>                       Generation type: image (default), thumbnail (YouTube-optimized), asset (logo/product)
+  --aspect-ratio <aspectRatio>        Aspect ratio (1:1, 16:9, 9:16, 4:3, 3:4)
+  --negative-prompt <negativePrompt>  Negative prompt
+  --seed <seed>                       Random seed for reproducibility
+  --reference-image <file>            Reference image file (auto-uploaded)
+  --wait                              Poll until generation completes
+  -o, --output <path>                 Download image to this path when done (implies --wait)
+  -h, --help                          display help for command
 ```
 
 ## image-edit
@@ -258,11 +242,12 @@ Usage: gobi media image-edit [options]
 Edit an existing image with a prompt (image-to-image).
 
 Options:
-  --media-id <mediaId>  Source image media ID
-  --prompt <prompt>     Edit instruction
-  --name <name>         Name for the edited image
-  --wait                Poll until generation completes
-  -h, --help            display help for command
+  --image <file>       Source image file (auto-uploaded)
+  --prompt <prompt>    Edit instruction
+  --name <name>        Name for the edited image (auto-generated if omitted)
+  --wait               Poll until generation completes
+  -o, --output <path>  Download image to this path when done (implies --wait)
+  -h, --help           display help for command
 ```
 
 ## image-inpaint
@@ -273,12 +258,13 @@ Usage: gobi media image-inpaint [options]
 Inpaint an image region using a mask.
 
 Options:
-  --media-id <mediaId>           Source image media ID
-  --mask-media-id <maskMediaId>  Mask image media ID
-  --prompt <prompt>              Inpainting prompt
-  --name <name>                  Name for the inpainted image
-  --wait                         Poll until generation completes
-  -h, --help                     display help for command
+  --image <file>       Source image file (auto-uploaded)
+  --mask <file>        Mask image file (auto-uploaded)
+  --prompt <prompt>    Inpainting prompt
+  --name <name>        Name for the inpainted image (auto-generated if omitted)
+  --wait               Poll until generation completes
+  -o, --output <path>  Download image to this path when done (implies --wait)
+  -h, --help           display help for command
 ```
 
 ## image-status