開発日誌

You can now run VlogMe from your own code and from AI agents. Every job the app starts — in Director, the Studios, Aisma, DUB and Trends — can be started by a request, with your account's credits and your Lab.

• Access (updated September 30, 2026): API and MCP are available to every registered VlogMe account. Create a key in Account → API & MCP; no support approval is needed. Volume and team integrations go to sales@vlogme.ai.
• Keys: create and revoke them in Account → API & MCP. A key is read-only or can also start jobs, and expires after 1 to 365 days.
• Jobs: upload a file (POST /api/assets/intent) or reuse one from Lab, get the exact credit price (POST /api/jobs/quote), start the job at that price (POST /api/jobs), then poll it or receive a signed webhook. Results are saved to Lab.
• MCP: connect an MCP client to https://vlogme.ai/mcp with the same key. Its tools read your account and balance, list and download Lab files, list and read jobs, quote a job and start it.

The API reference has the requests, the webhook signature and the limits. This replaces the first public API announced on June 6, 2026.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 1 smaller or behind-the-scenes improvement.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 15 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 21 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 20 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 32 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 21 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 37 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 3 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 35 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 35 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 8 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 13 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 20 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 37 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 14 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 6 smaller or behind-the-scenes improvements.

A concise daily roundup of shipped changes from the VlogMe team.

Plus 4 smaller or behind-the-scenes improvements.

Create projects can now place text directly on video with preview-to-output positioning that matches. We also fixed cloning and editing paths that could crop, downscale, hide, or overwrite the wrong media.

We fixed public-page behavior in Safari and improved Studio media refresh when a signed Storage link expires. The result is fewer blank previews and fewer dead ends on long-open pages.

Before sign-in moved to the new system, we added a focused recovery experience for existing users and temporary read-only access to their media during the move. This reduced the chance of users getting stranded.

Uploads and drafts started on public tools now survive the sign-in handoff more reliably. We also improved funnel diagnostics so broken transitions can be found without exposing user content.

We added dedicated model pages, richer model discovery, and a Replicate-backed Seedance option. Image and Video Studio now make it easier to understand model strengths, pricing, and the right workflow before generating.

The Dashboard and Lab were rebuilt around real motion examples, faster capability discovery, richer media metadata, and responsive layouts. It is now easier to start from inspiration and jump into the right Studio.

We rebuilt the homepage around real examples, clearer workflows, and localized visual proof instead of abstract feature claims. Visitors can now understand what VlogMe creates before opening the Studio.

You no longer have to wait for one Image, Video, or Audio Studio job to finish before starting another. Submission and progress handling were redesigned so multiple creative jobs can move in parallel.

We added a clearer path from a source image to video creation and made checkout resume the interrupted task safely after payment. Pricing and authenticated handoff now preserve more of the creator's context.

The Dashboard gained daily quests, streak-aware progress, reminders, and safer activity reconciliation. The coach connects creation, publishing, and follow-up without inventing progress that did not happen.

We fixed stale draft and tutorial state that could leak into a new guided session. Ending or restarting the tour now preserves real work while clearing only the temporary tutorial context.

Completion notifications for Studio and Create jobs now open the finished media in Lab. Delivery is recorded once, so retries do not turn one finished video into repeated alerts.

Update, September 29, 2026: this post describes the first version of the API documentation. The current API reference documents the job-based REST API and the MCP server, with curl examples for quotes, jobs, uploads and webhooks.

API documentation gained copy-paste curl, Python, and JavaScript paths from token creation through render completion. It also documents estimates, idempotency, webhooks, errors, and typed OpenAPI clients.

We restored semantic learning from accepted and edited scripts, repaired honest grading, and added self-healing for missing embeddings. The Director can now learn from what users actually render instead of treating every draft as a perfect answer.

When post-processing cannot finish, its reserved credits are refunded through an idempotent billing path. Retried failure callbacks stay safe and cannot return the same credits twice.

Projects can now be authored, previewed, rendered, and exported in horizontal and square formats as well as vertical. The format choice flows through scripts, media picking, canvas previews, provider routing, and API documentation.

We expanded speech-to-speech, audio isolation, AI image editing, provider controls, and Studio render paths in one coordinated update. Creative tools became more capable and more consistent across workflows.

Video editing jobs from Library were connected to the worker fleet with explicit API contracts and routing. Larger edits can now run away from the request path and complete more reliably.

We introduced render-fleet autoscaling, worker discovery, and capacity-aware scheduling. Rendering infrastructure can now grow with demand while keeping the application responsive.

Avatar rendering was moved behind a dedicated Worker API with clearer provider routing and preview handling. This created a safer foundation for scaling renders independently from the web application.

A few quieter fixes that make a big difference once you notice them.

What's fixed

• Subtitles no longer drift on scenes with Hunyuan inserts. We now snap the insert length to Hunyuan's real 24fps / (4n+1)-frame grid, and if the insert is longer than the avatar's speech we trim it to end exactly with the avatar — so subtitles stay perfectly in sync from the first word to the last
• Every voice block now gets a [short pause] lead-in before being sent to ElevenLabs — no more clipped first syllables or jarring starts
• Emotion tags like [happy], [laughs], [pause] in your script no longer trigger the "this looks like a prompt" warning — they're recognized as performance cues

Nothing for you to enable — these apply to every new render. If you still see drift on a finished video, hit Re-create and it'll come out clean. 🎬

Drafts used to be a mess — duplicates everywhere, clicking a draft opened a read-only video page, you'd lose work if you closed a tab too early. Fixed.

What's new

• The moment you drop the first photo on /create, your project is autosaved as a draft automatically
• Every edit after that — script, voice, music, settings, inserts — updates the same draft (no more 14 copies in My Videos)
• Draft cards in My Videos now open straight into the editor (/create), not the read-only video page — so you can actually keep working on them
• Hitting Create promotes the existing draft into a render instead of creating a brand-new row

Close your tab without worry. Pick it up tomorrow. It'll be there exactly as you left it. ✨

Picking what to show on the home page used to be pure gut feeling. Now AI helps.

What's new

• Every featured example is auto-scored by Gemini for viral potential — image energy (about 55%) and script humor / surprise / pacing (about 45%)
• Scoring uses a fine 10–1000 scale, so two strong examples don't end up in the same bucket
• Higher-rated examples surface more often via a weighted "popularity boost" that runs every 3 minutes — strong videos rise faster, weak ones don't get stuck at the bottom
• New uploads land in the gallery first, get a rating, and only the best get promoted to the home page with one admin toggle

Result: the home strip stays fresh and the gallery stays browsable. Comments and counter-examples welcome. 👇

We just launched a real /gallery page — every published example in one place, with the kind of buttons people actually wanted. ❤️

What's new

• Full gallery route with every featured example, not just the home strip
• ❤️ Like button on home and gallery cards (with instant counter)
• Open project — clones the whole example into your drafts: scenes, voices, overlays, music, prompts — not just the poster
• Sort by Most liked or Newest, with an "Added Xd ago" label on each card
• The home strip is now the curated list — admins pick which examples appear and in what order; the rest live in the gallery
• Behind the scenes: every gallery source is archived independently, so a published example never breaks if the original creator deletes their project

Open /gallery, give a few hearts, then hit Open project to remix one. We'd love to hear what you build. 🎬

The Music dialog finally has music in it. 🎵

What's new

• 16 ready-to-use presets across every vibe: Lo-Fi, Pop, Cinematic, Corporate, EDM, Acoustic, Hip-Hop, Deep House, Rock, Ambient, Jazz, Epic Trailer, Synthwave, Tropical, Tech, Fashion
• Brand-new Generate music with AI panel right inside the dialog — describe the vibe, pick a length 10–300 seconds
• Every track you generate is saved to your personal music library for reuse on future videos
• Fixed: the Music button now opens the dialog whether watermark is on or off (it used to silently no-op for some folks)

Open /create, hit Music, and either pick a preset or describe your own. Let us know what genres you'd like added next. 🎧

When your video is done, the next action is almost always publish it — so we made publishing the obvious next step.

What's new

• On the finished-video page, Publish is now the primary CTA (with Instagram / TikTok / YouTube icons stacked on the button)
• Download MP4 is now secondary, Re-create stays right next to it — same actions, cleaner hierarchy
• Delete moved to the bottom of the page so you stop hitting it by accident
• Whatever social platforms you tick last is remembered everywhere you publish — main video page, ad-hoc card, modals — and syncs live across tabs

Less clicking, fewer regrettable deletes, faster shipping. 🚀

We just shipped the Visual Builder — a brand new way to put together amazing short videos in just a few clicks.

What's new

• Drag-and-drop timeline: arrange scenes, inserts and voiceovers visually
• One-click scene creation from a photo + short script
• Smart presets for voice, language and aspect ratio carried across scenes
• Live preview while you build, no guessing how the final clip will look
• Re-run the in-app tutorial any time from the Coach button on /create

Open /create and give it a try — we'd love to hear what you think in the comments. 🎬

AI Ideas = an infinite, personalized feed of short-form video ideas. Stop staring at a blank editor.

What it does

• ♾️ Scroll an endless stream of ideas tuned to YOUR niche, audience & tone
• 🧠 Each idea = title + hook + angle + suggested format (talking head / listicle / story / POV)
• ❤️ Like / 👎 dismiss — the feed learns and gets sharper every scroll
• 📌 Save for later → goes to your idea bank
• ⚡ One tap → AI Director turns the idea into a full script & scene plan, ready to render

Example (niche: SaaS founder)

💡 "3 lies your landing page is telling investors" — listicle, 35s, bold claim hook 💡 "POV: you raised $500k and your churn is 12%" — story, 25s, vulnerable open 💡 "I A/B tested 7 pricing pages so you don't have to" — listicle, 45s, curiosity 💡 "Why 'book a demo' is killing your funnel" — contrarian, 30s

Tap any → AI Director drafts the script → pick voice/avatar → render. ~2 min, idea to video.

How /create or Dashboard → AI Ideas feed → scroll, like, save, or tap to make.

Fill in Settings → AI profile to make the feed obsessively on-brand.

— Developer

Trend Radar = what's actually winning on TikTok / Reels / Shorts right now, filtered to your niche.

What it does

• 📡 Live trending hooks, topics & sounds (refreshed all day)
• 🎯 Auto-filtered by your niche & audience
• 📈 Momentum tag: Rising / Peaking / Cooling — never jump in late
• 💡 "Why it works" one-liner so you adapt instead of copy
• ⚡ One tap → AI Director writes a script in YOUR voice on that trend

Example

Niche: fitness coach

🔥 RISING — "POV: I tried [X] for 30 days" (curiosity + payoff) 📈 PEAKING — "Stop doing [common advice]" (contrarian) ❄️ COOLING — sound: "Murder on the Dancefloor" remix

Tap → AI Director: "POV: I tried zero cardio for 30 days. Here's what happened to my abs…" — full script, ready to render.

How Dashboard or /create → Trend Radar card → tap a trend → Use this → tweak → render.

— Developer

AI Director = your in-app creative co-pilot. From blank page to a renderable short in ~60 seconds.

What it does

• 🎣 Hooks in 3 styles (curiosity / bold claim / story-open)
• 📝 Full 20–45s script in your voice
• 🎬 Auto scene breakdown + B-roll suggestions
• 🎤 Picks the best voice & avatar for the vibe
• 🔥 Title + on-screen text variants
• ✂️ Inline rewrites: "punchier", "shorter", "more emotional", "Gen Z it"

Example

You: "3 mistakes new founders make on Twitter"

AI Director: Hook: "I lost 6 months on Twitter doing this — don't." Script: 38s, 4 scenes, CTA "Follow for the fix" Voice: Marcus (confident, mid-tempo) Title: "Don't post on X until you fix this"

How /create → AI Director → drop a topic → pick tone → render.

More signal in Settings → AI profile = sharper output.

— Developer

You can now add cinematic special effects directly inside your script. No timelines, no extra editors — just write what you want to happen in {curly braces} and we'll render it.

Mix effects with @image tags freely:

@image1 Hey, how's it going? {camera slowly zooms in} @image2 Watch this. {a meteor crashes behind me, dust and sparks fly} @image1 Whoa. {everything goes black and white for a second}

Anything you describe inside { } becomes a visual effect in that exact moment — explosions, weather, lighting changes, camera moves, transitions, magic, you name it. Combine it with multi-shot @image1 / @image2 scenes for full mini-movies from a single prompt.

Open Create and try it on your next video.

You can now build multi-shot scenes inside a single generation — insert as many photos as you want and tell us, in plain text, who speaks in what order.

Just drop your images into the scene and use simple tags in your script:

@image1 Hi everyone, today we're testing a new product. @image2 And here's what my friend thinks about it. @image1 Let me show you the result.

The avatar swaps between your photos exactly in the order of the tags, so you can build full dialogues, product demos, reactions, and back-and-forth conversations from one prompt — no stitching, no separate renders.

Try it now in Create → add multiple portraits → reference them as @image1, @image2, @image3… in your script.

We shipped something big: you can now chain multiple scenes in one generation, with different characters and different voices speaking in turn.

What this unlocks:

• Mini-stories and short skits with several speakers
• Dialogues and interviews between characters
• Scene-by-scene narratives — intro, twist, punchline — rendered in one go
• Each scene can use its own portrait, voice, and script line

It's basically a multi-shot mode: instead of one talking head per video, you direct a small cast.

We're still naming the feature. Should we call it Multi-scene, Multi-shot, Story Mode, or something else? Drop your ideas in the comments 👇

Hey everyone — quick update from the dev team.

We just shipped a big fix to the singing pipeline. Your avatars now actually sing — with proper pitch, timing and a lot more emotion. In some takes it might even feel a little too emotional 😄

Give it a try on your next video and let us know:

• Does the emotion feel right for your style?
• Any songs or voices where it goes over the top?
• Anything else you'd like us to tune?

Drop your feedback and wishes in the comments — we're reading everything and iterating fast. Thanks for pushing us!

Update, September 28, 2026: the program has no cookie and no dashboard. A partner link (vlogme.ai/?partner=YOURCODE) is remembered in the visitor's browser for up to 30 days, and a partner code can be typed on the sign-in page; either way the new account is recorded as your referral. Commissions clear after 30 days, and available balances of $50 or more are paid monthly. The rate is 20%, as announced. The affiliate program terms have the details.

We just launched the VlogMe Affiliate Program — and it's built to actually pay creators.

What you get

• 💰 20% of every payment your referrals make in their first 12 months
• 🍪 60-day cookie window (replaced: see the update above)
• ⚡ Real-time dashboard with clicks, signups, paying users and earnings (not offered: see the update above)
• 💸 Payouts once you cross the threshold

How it works

1. Apply by email from /affiliate/apply
2. We review and approve you
3. Share your personal partner link (its current form is in the update above)
4. Earn 20% of their payments for 12 months

Whether you have a YouTube channel, a newsletter, a Discord, or just love VlogMe — this is a no-brainer way to turn it into income.

👉 Apply to the affiliate program

Questions? Drop them in the comments.

Update, September 28, 2026: this post describes the first version of the API. Its /api/public/v1 endpoints and the Settings → API page were replaced by the job-based REST API and the MCP server at /mcp, with keys in Account → API & MCP: see the API reference and the announcement of September 28, 2026.

Big update for everyone living in Claude, Cursor, Windsurf and other AI agents.

vlogme.ai now speaks MCP (Model Context Protocol) and ships a full REST API. You can generate talking-head videos directly from your agent — no UI, no copy-paste.

What you get: • MCP server — plug vlogme into Claude Desktop, Cursor or any MCP-compatible client. Your agent can list voices, pick an avatar, write a script and render a video in one flow. • REST API v1 — /api/public/v1/videos, /voices, /me. Build your own integrations, n8n flows, cron jobs, Telegram bots, whatever. • Personal API tokens — generate and revoke them in Settings → API. • Same credits, same pipeline — anything you can do in the web app, you can now do from code.

Docs: /docs/api Tokens: Settings → API

We'd love to see what you build. Drop your demos and ideas in the comments.