Ad Agent is now live on Teno: an AI agent that turns a chat into finished image and video ads. You describe the product, goal, and channel. The agent structures an ad brief, auto-starts generation with no Confirm button, and runs built-in quality checks that retry weak image runs up to 3 times per generation (Teno product limits, June 2026). It is the campaign-focused counterpart to Asset Generator: not a model studio, but an agent that drives the whole brief-to-ad loop for you.
Key Takeaways
- Chat in, ads out. Ad Agent collects your goal, channel, style, and format through conversation, structures a brief, and generates the ads, with no prompt engineering or model picker.
- Auto-generate with built-in QA. Once the brief is clear, generation starts automatically. A vision model grades each image run against the brief and retries up to 3 times (1 initial submit + 2 extra) before settling (Teno product limits, June 2026).
- Image and video in one thread. Static image ads and composed multi-scene video ads with voiceover up to 60 seconds are separate deliverables inside the same chat.
- Free to start, charged per step. Image runs show an upfront credit estimate; video charges per scene as it completes. All on Teno's free monthly credits with no card required.
This post is the canonical Teno write-up on Ad Agent: what an AI ad agent is, how the chat-to-brief-to-ad loop works, how the built-in QA retry behaves, how it differs from Asset Generator, and what it costs. For the wider category framework Ad Agent sits inside, see our AI ad creative platform guide.
What is an AI ad agent, and how is Ad Agent on Teno different?
An AI ad agent is software that takes a campaign intent in plain language and drives the full creative loop: brief, generation, and quality control. It does that instead of leaving you to assemble prompts and pick models yourself. Platform-native automation sits alongside this workflow. Meta's Advantage+ creative tools, for example, diversify and enhance assets at delivery time. Campaigns using its image-generation feature report 11% higher CTR and 7.6% higher CVR on average (Meta for Business, December 2025). An advertiser-side agent like Ad Agent owns the brief-to-asset loop before upload. Ad Agent on Teno is the ad-creative version of that idea: one chat thread that ends in downloadable image or video ads. The brief and the image QA loop are handled for you.
In one sentence: Ad Agent is Teno's campaign-focused AI ad agent that turns a chat brief into finished image or video ads, then runs built-in image QA before surfacing the result (Teno product limits, June 2026).
For marketers, an AI ad agent is best defined narrowly: it accepts campaign intent, turns that intent into an executable brief, produces the creative, and checks output quality before handoff. A tool that only generates images from prompts is better described as an AI ad generator, not an agent.
The design choices that separate it from a general creative tool:
- The brief is the interface, not a prompt box. You talk about the product, audience, channel, and look. The agent turns that into a structured ad brief (goal, style, format, deliverable) and shows it back to you. You never write a model prompt.
- Generation auto-starts. There is no "Confirm" gate. When the brief is clear enough, the run begins immediately. Keep chatting and a clearer brief triggers a new run in the same thread.
- Quality control is built in. A vision model evaluates each image run against the brief and retries weak results automatically, so you are not the first reviewer of a bad generation.
Ad Agent is purpose-built for marketers shipping campaign creative, not a general-purpose sandbox. For the open studio where you choose among curated image and video models yourself, that is Asset Generator; the two share one account and one credit pool.
How does Ad Agent turn a chat into a finished brief?
Ad Agent collects intent through conversation, then commits it to a structured brief with four fields the generation step actually needs: goal, style, format, and deliverable type (Teno product limits, June 2026). The model orchestration stays behind Teno's product interface, so the back-and-forth feels like briefing a junior creative, not filling a form.
The flow in practice:
- You describe the campaign. "Sparkling water summer launch for YouTube: a short video ad with voiceover, fresh studio light, crisp premium mood." Attach a product photo if you want the real product to stay consistent.
- The agent proposes a brief. It fills the gaps it can infer (aspect ratio, scene count) and asks about the ones it cannot. The brief renders as a read-only card in the thread.
- Generation starts on its own. No Confirm button. A clear brief is the trigger; the run begins and streams progress back into the chat.
Because the brief lives in the conversation, refining is just talking. Change the channel from YouTube to TikTok, swap the mood, or bump the scene count. Each clear revision proposes a new brief and a new run. Earlier runs stay in the thread so you can compare.
How does built-in QA work?
Once a brief is clear, Ad Agent starts a generation run and then grades its own output. A vision model evaluates each image set against the brief and patches weak runs before retrying, up to 3 image submits per trigger (1 initial + 2 extra; Teno product limits, June 2026). The QA loop runs in Teno's durable generation pipeline. An interrupted run can be tracked, retried, or refunded instead of silently burning credits.
What the QA step checks and does:
- Evaluate. A vision model scores the generated images against the structured brief (goal, style, format alignment).
- Retry with a patch. If a run is weak, the agent adjusts safe brief and instruction fields while keeping your frozen product reference unchanged, then regenerates.
- Settle. After at most 3 submits, the best run lands in the thread with a QA trace you can inspect.
For AI ad workflows, the important constraint is the retry cap: image QA can submit up to 3 attempts per trigger, then returns the best available run with an inspectable trace. That limit keeps the automation bounded while still filtering weak first drafts before the user reviews them.
Our finding: the built-in retry is the single biggest difference between Ad Agent and a one-shot generator. In the QA card above, the accepted image scores 99 against the brief before it reaches the thread. A plain text-to-image tool hands you whatever the first sample produced. Ad Agent treats the first sample as a draft to be graded, and only surfaces a run after it has passed (or exhausted) the QA gate.
So why is there no Confirm button? A Confirm gate usually protects against wasting a generation on a misread brief. Ad Agent handles that with the evaluate-and-retry loop instead of a human click. The check still happens. A machine just does it faster.
Can Ad Agent make video ads, not just static images?
Yes. Ask for a video and Ad Agent plans a multi-scene storyboard with voiceover, generates each scene, and composes them into one video ad up to 60 seconds (max 10 segments at roughly 6 seconds each; default 20 seconds across 3 scenes at 9:16, 720p; Teno product limits, June 2026). Image and video are separate deliverables: you pick one per brief, and you can run both in the same chat thread.
How a video run is built:
- Storyboard. The agent breaks the brief into scenes, each with a keyframe and a voiceover line.
- Per-scene generation. Each scene is produced as an image-to-video clip, so the product and look stay consistent across cuts.
- Voiceover and compose. The agent generates narration (multilingual, with Chinese routed to a dedicated voice model), times it to each scene, and composes the final cut.
Ad Agent is best described as a chat-to-video-ad workflow, not a raw video model. It plans scenes, generates clips, adds voiceover, and composes the final ad; the underlying model orchestration stays behind Teno's product interface.
The default video is short on purpose: a 20-second, 3-scene cut covers most social and pre-roll placements. When a brief calls for more, the agent scales up to the 60-second cap. Attach a product reference and the image-to-image path keeps your real product on screen across every scene, instead of inventing a look-alike. The broader image-to-video model category is moving quickly too; ByteDance Seed's Seedance 2.0 launch notes are a useful reference for why scene-by-scene ad generation is becoming practical. For how Teno's brief-led video path stacks up against other tools, see our roundup of the best AI video ad generators.
When should you use Ad Agent instead of Asset Generator?
Both are first-party Teno surfaces that share one account and one credit pool, but they solve different jobs. Ad Agent is the agentic, campaign-first path: chat to brief to ad, with QA built in. Asset Generator is the studio: you pick from 14 curated image and video models and drive the iteration yourself (see the Asset Generator launch post for the full model catalog).
| Dimension | Ad Agent | Asset Generator |
|---|---|---|
| Primary job | Campaign ads from a chat brief | Open-ended image and video creation |
| Model choice | Handled for you (server-selected) | You pick from 14 curated models |
| Starting point | Conversation → structured brief | A prompt and a model |
| Quality control | Built-in vision-model QA with retries | You review and regenerate manually |
| Iteration unit | Brief revisions in one thread | Prompt and model swaps in one thread |
| Best fit | Marketers who want finished ads, fast | Creatives who want full model control |
Best-fit rule: use Ad Agent when the campaign objective matters more than model selection, and use Asset Generator when the model choice or generation parameters are the main creative decision. That distinction is the simplest way to separate an AI ad agent from a general AI media studio on Teno.
Where Ad Agent is the wrong tool: if you want to compare the same prompt across several models, hand-tune generation parameters, or make non-ad creative, Asset Generator gives you that control and Ad Agent deliberately hides it. The honest one-liner: Ad Agent decides for you so you can ship; Asset Generator hands you the dials. Many teams use both. Ad Agent handles fast campaign output; Asset Generator steps in when a specific shot needs a specific model.
How are credits charged for an Ad Agent run?
Ad Agent is included in Teno's free tier with monthly credits and no credit card required. Image runs show an upfront credit estimate on the brief card, and video runs charge per step as each scene completes, so there is no single upfront total for a full video generation (Teno product limits, June 2026). Built-in QA retries deduct per image attempt, capped at 3 submits per trigger.
The cost gap between a static image set and a video ad is large, which shapes how teams use the tool:
A practical read: a single still costs 1 credit, a 3-image variant set costs 3, and a default 20-second 3-scene video runs about 30 credits in per-scene generation alone (voiceover and compose are computed but not billed today).
The practical billing takeaway is that still-image exploration is the cheapest way to validate a concept before spending credits on video. Because video billing happens scene by scene, teams can use image variants to lock the look, then generate the 20-second default video when the direction is clear. For live credit-to-USD conversion and higher monthly allocations, see pricing.
How do you iterate when an ad is not right yet?
You keep talking. Iteration in Ad Agent happens in the same chat thread: adjust the brief in plain language, change the style, format, message, or channel, and a clear revision triggers a new run while earlier runs stay visible for comparison (Teno product limits, June 2026). There is no separate "edit" mode to learn.
That matters because performance teams need creative variety, not one perfect file. TikTok for Business recommends keeping 5–7 varied creatives in rotation and refreshing weekly for performance campaigns (TikTok for Business, retrieved 2026-06-08). Ad Agent's job is not to bet everything on one asset; it lowers the cost of getting to the next variant.
Two refinement paths matter most:
- Refine the brief. "Make it warmer and add the price callout." The agent proposes an updated brief and regenerates.
- Anchor on your product. Attach a product photo and the run switches to image-to-image, preserving the real product across the scene instead of redrawing it.
Because every run is preserved in the thread, the workspace doubles as a comparison view. You download the version you like best; the rejected drafts stay as context for the next revision rather than disappearing.
Frequently asked questions
How is Ad Agent different from Asset Generator?
Asset Generator is a flexible studio for general image and video creation where you pick the model. Ad Agent is campaign-focused: it structures an ad brief from your chat, auto-starts generation, and runs built-in image quality checks so the output is ad-ready without prompt wrangling.
Do I need to approve the brief before generation starts?
No. When the agent has a clear brief, generation starts automatically; there is no Confirm button. You keep chatting to refine the goal, style, or format, and each clear new brief triggers a fresh run in the same thread.
Can I attach a product reference image?
Yes. Add a product photo in the composer. When the brief includes a reference, Ad Agent uses image-to-image generation to preserve your product appearance inside the ad scene, instead of redrawing the product from a text prompt alone.
Does Ad Agent support video ads?
Yes. Ask for a video in chat and the agent plans the scenes and voiceover, then composes them into one video ad up to 60 seconds. Image and video are separate deliverables; pick one per brief, and run both in the same chat thread.
How are credits charged?
Image runs show an upfront estimate on the brief, and built-in quality checks may run up to 3 image attempts per trigger. Video ads charge per step as each scene completes, so there is no single upfront total for the full video generation. See pricing for live conversion.
Try Ad Agent
Prefer to watch first? See the 1-minute product walkthrough on YouTube on the Teno channel.
Ad Agent is live today on every Teno plan, with the free tier open and no credit card required. The fastest test: start a Teno account, open Ad Agent, and describe one campaign like "summer launch for a sparkling water, fresh studio look, for YouTube." Then watch it brief, generate, and QA a set of image ads, and ask for the video once you like the look.
For the feature overview, see the Ad Agent page; for the open model studio, see Asset Generator; and for the wider category framework Ad Agent sits inside, see the AI ad creative platform guide.
Sources
- Teno Ad Agent product limits and pricing, June 2026 (built-in QA capped at 3 image submits per trigger; video defaults of 20 seconds across 3 scenes and the 60-second / 10-segment cap; default 9:16, 720p; per-step video billing with voiceover and compose computed but not billed). Live credit-to-USD conversion reflected on /pricing.
- Meta for Business, Demystifying Creative Diversification, retrieved 2026-06-08 (Advantage+ creative automation and generative image/text uplift benchmarks).
- TikTok for Business, Return On Influence: How Creators and Creative Variety Can Spark Performance, retrieved 2026-06-08 (5–7 varied creatives and weekly refresh benchmarks for performance campaigns).
- ByteDance Seed, Official launch of Seedance 2.0, retrieved 2026-06-07 (image-to-video model family context for the video pipeline).
