Why Can't My AI Girlfriend Send Video? Locate the Failed Gate

Locate the failure before rewriting the prompt

Rewriting the request before locating the failure can waste time. The first diagnostic step is to identify which gate has evidence and which remains unknown. Video delivery is not a single feature but a pipeline with sequential checkpoints. Different failed checkpoints can produce the same visible symptom: no video appears.

A structured diagnosis prevents chasing symptoms that cannot be fixed by better wording. For example, if the product simply lacks any video pipeline, no amount of prompting will create one. If the specific motion requested has no pre-generated asset, the system cannot invent it. Locating the exact failure point lets you decide whether to adjust the request, upgrade, switch products, or stop.

Capability, entitlement and asset supply are separate gates

Capability, entitlement and asset supply operate as independent controls. Conflating them leads to incorrect conclusions.

Capability asks whether the product can deliver any video at all. Some companions have no video generation or gallery system. Others support only still images or voice. Kindroid documents generated video selfies and separate live avatar calls. Replika lists background calls and tiered selfie/video features. AISoul supplies short asynchronous clips drawn from a pre-generated companion gallery; it does not perform live calls or generate new video from each prompt.

Entitlement determines whether your account level may retrieve video. Free accounts on AISoul receive 2 clear clips lifetime plus daily allowances of 50 messages and 5 gallery photos that reset by Beijing calendar day. Paid passes unlock eligible 18+ media and remove the published quantity cap on eligible gallery retrieval while active. Prices are 7-Day $4.99, 30-Day $8.99, 90-Day $19.99 and Annual $49.99, each a fixed-duration one-time Pass with no automatic renewal. Payment method, tax, FX and bank approval vary. Entitlement is not the same as asset existence.

Asset supply checks whether a suitable clip actually exists in the finite gallery. The system selects the closest eligible match or a fallback using the request and media descriptions. The accompanying text reply should follow the actual asset supplied. A repeated request cannot force a nonexistent clip into existence. This is not prompt-to-video generation.

Check capability and entitlement first, then compare any returned asset with the request. A user-facing interface may not expose the full gallery, so treat supply as unresolved unless documentation or support confirms it.

Translate five visible symptoms into the next test

The same “no video” complaint can start at different gates. Use one row, make its single controlled check, and stop before moving to the next explanation.

Completed example: an AISoul free account has already received two clear clips, shows no active paid Pass, and now returns text without a third clip. The evidence matches the published lifetime allowance. The next choice is to buy a Pass or stop requesting clips; waiting for the daily message/photo reset cannot restore the lifetime clip quota.

Match motion requests to the product's video format

AISoul does not generate video live from prompts. It selects from a finite library of pre-generated clips. Therefore the request must describe motion that plausibly exists in that library.

For one reader-run test, describe mood, framing, clothing level and action without adding conflicting instructions. This does not guarantee a match; it simply makes the request easier to compare with the returned asset. Because AISoul's reply is intended to follow the actual asset, a disagreement between text and clip belongs in the evidence packet.

Compare your request against the three internal format boundaries:

- Short asynchronous clips versus live video calls: /research/ai-video-clips-vs-video-call-girlfriend.html

- Realistic adult-chat expectations for photos versus video: /research/adult-ai-chat-expectations-photos-video.html

- Current 2026 landscape of companions that actually deliver video: /research/best-ai-companion-with-video-2026.html

If the desired motion falls outside the product’s documented capabilities, the gate is format mismatch. No prompt rewrite will create a new format.

Escalate delivery failures with an evidence packet

When documentation confirms capability, the account confirms entitlement, and an allowed request still produces text without an attached file, delivery becomes the next candidate to test. The visible result does not identify a network, server or asset-index cause.

Compile a compact support packet: exact request, timestamp, screenshot of the placeholder or error, current tier, date of the last successful clip and the product documentation link. This lets support compare the promised format with the observed result without guessing which prompt was used.

Do not label the whole product broken from one missing file. Record the event, make at most one controlled retry after checking account state, and escalate only the repeated result.

Stop retrying when the requested medium does not exist

Continued prompting after documentation or support confirms that the requested format is absent wastes quota. In a finite-gallery product, another request may return a close match or fallback but cannot turn the service into on-demand video generation.

Recognize only a verified boundary: stop when official documentation or support confirms that the format is absent, the free lifetime allowance is exhausted, or the requested category is outside current policy. Repeated mismatches alone show that the request has not been fulfilled; they do not prove what the unseen gallery contains.

This decision tool prevents sunk-cost repetition. Document the tested phrases and outcomes in a personal worksheet so future requests stay within verified boundaries.

Video-failure questions that change the diagnosis

Does the product you are using document any form of in-chat video clips at all?

Look for the exact format in the current feature or pricing page: short clip, generated video, animated avatar or live call. If only text and photos are documented, prompt rewriting cannot create a video pipeline. AISoul documents gallery short clips, not live calls or on-demand video generation.

Have you confirmed whether your account still has remaining lifetime clear-clip allowance?

On AISoul, free accounts have two clear clips over the account lifetime, not two per day. Check prior deliveries and the current upgrade state. Waiting for the Beijing daily reset can restore daily messages and photos, but it does not restore that lifetime clip allowance.

Is the motion you requested already represented by an existing asset in the companion’s gallery?

The public interface may not expose a complete asset inventory. Compare the returned clip with the requested action and ask support about availability if exact motion is essential. A mismatch proves only that the delivered asset did not match; it does not prove that a hidden matching clip exists.

Are you requesting short asynchronous clips or expecting a live video call?

Write the required format before troubleshooting. A short file delivered into a thread cannot provide real-time turn-taking, camera control or a live response. If a live call is the requirement, reject a gallery-clip product even when its clip delivery is working correctly.

Has the same request succeeded previously with this exact companion and tier?

A prior success confirms that some delivery path worked under those earlier conditions. It does not prove current entitlement, asset availability or service health. Record companion, tier, time and returned format, then change only one of those variables for the next test.

Does the failure occur only on explicit content or across all video requests?

Test one policy-safe motion request and one allowed adult request without changing companion or tier. Failure across both points toward capability, entitlement or delivery checks; a difference between them makes policy or eligible-supply checks more relevant. The comparison narrows the route but does not reveal the hidden cause.

Did you receive a text description of a video without an attached file?

Classify that as text-without-media, not a successful video response. Save the message, timestamp and network result if available, then retry once. If the same split repeats, send the evidence packet to support instead of rewarding the failure with more prompt variations.

Are you on a free daily reset plan or an active paid pass?

This decides which allowance to inspect. AISoul's free message and photo quotas reset by Beijing calendar day, while the two clear clips are lifetime. An active paid Pass removes the published quantity cap on eligible clip retrieval, but still cannot promise a unique or exact-scene file.

Video mechanisms, verified allowances and unmeasured performance

Exact performance characteristics such as delivery latency, clip-length distribution and match accuracy were not independently measured. The boundaries above come from official documentation; this article does not add an unpublished performance claim.

Official sources

- AISoul pricing and allowance details: https://www.aisoul.work/pricing.html

- AISoul about and technical approach: https://www.aisoul.work/about.html

- AISoul privacy and processing disclosures: https://www.aisoul.work/privacy.html

- AISoul terms of service: https://www.aisoul.work/terms.html

- Kindroid video documentation: https://kindroid.ai/v2/docs/selfies-video-selfies-avatars/

- Replika subscription features: https://help.replika.com/hc/en-us/articles/39551043419149-Choosing-a-Subscription

Update note: Allowance, pricing and mechanism facts were verified 2026-09-07. Re-check the official pages after a product update; the table distinguishes documented format from performance this review did not test.